# Theories Explaining Relationship Of Stock To Dividend Changes Finance Essay

Jules Regnault first introduced the concept of random walk that the prices in stock market move randomly and cannot be predicted. The concept was further enhanced by Louis Bachelier (1900) in his paper 'the theory of speculation'. Jones and Cowles (1937) stated that investors are unable to outperform markets until they have some unique information that others in the market do not have. Eugene Fama (1970) in his article 'efficient capital markets' introduced the concept of efficient market hypothesis and since then it has been argued by researchers and academicians that market price reflects all the information and investors cannot make abnormal profits unless they have some private information which is not available publicly to other investors. The topic of how some announcements and actions of managers, having inside information, effects stock price has been researched since 1950s. Semi-strong market efficiency hypothesis has been tested using various studies in which the impact of announcement of information such as earning announcement, dividend announcement, stock repurchase announcement and mergers announcement has been extensively investigated. Dividends and earnings are the most common announcements made by the companies and are used more often than others. Earnings are considered to be reflecting current performance while dividends are believed to be providing information regarding future performance expectations by the management.

Most common method of payment of dividends is cash dividends and is believed to provide signals to the investors in the market. There is information asymmetry; the management being in the company has access to superior information about the current and future financial position of the company compared to information that is with the investors. Investors then use announcements by the management of the companies to get signals about the future performance of the company (Daniels, Shin, and Lee, 1997). Therefore the announcements made by management specially the declaration of cash dividends convey valuable information to investors about the management's future expectations about the financial performance of the company. Consequently, the cash dividend increase is considered as positive news and usually cause increase in the stock price while cash dividend decrease is considered as negative news and is usually followed by decrease in stock price. This leads to the 'information contents of the dividends hypothesis' or 'dividend signaling hypothesis' proposed by Linter (1956) which is further researched and developed by Fama, Fisher and Jensen (1969) and John and Williams (1987). Contrary to this, Modigliani and Miller (1961) have stated that dividends are irrelevant for investors as they do not have any effect on prices in world where there are no taxes and no transaction costs. However, previous research has findings that support the argument that announcement of dividend does in reality have an effect on the stock prices.

Cash dividend is considered as payment to shareholders from what they own hence the cash dividends paid are equal to the fall in the value of the stock price (Porterfield 1959 & 1965). This idea of relevance of dividends was further enhanced by Walter (1956) & Gordon (1962) which was then formalized as a theory of dividend relevance. It proposes that current fair value of the stock is equal to the sum of present values of its expected future dividends. Whereas Penman (1983) suggested that management's forecast on earnings has more information content as compared to dividend announcements and is better indicator of company's value.

Dividends can be paid in different forms including cash, stock bonus, property stocks, and convertible stocks. This study focuses on assessment of whether the cash dividend announcement has any effect on the price of the stocks in the market. A dividend payment is source of cash flow to the shareholders but at the same time for the company it is a decrease in cash available for investments. Modigliani and Miller (1958) stated that since dividends are irrelevant for the investors hence these dividends should be invested in the positive net present value projects to generate more value for the shareholders. Literature also suggests that the price of the stock depends on the expected future dividends from the company. Companies pay dividend out of what it earns and if it pays its earnings out it means that future investments will decrease leading to decrease in future earnings and with decreased earnings less is available to pay as dividends. Moreover, if dividends are taxable, paying out higher dividends will increase shareholder's tax liability. Despite these orthodox arguments for not paying dividends companies still pay dividends possibly to signal information about future performance of the company. There may be different signals associated with dividends. Increase in dividend might be taken as that company has made higher profits and have abundant earnings available or that it has excessive cash which it is distributing because it no more has value adding projects to invest in. Similarly, decrease in dividends may convey the message that company has underperformed and did not made sufficient profits to be distributed or company has a lot of value adding opportunities where it want to invest instead of paying out as dividends.

Karachi stock exchange, established in 1947, is first, largest and most important stock exchange market in Pakistan, which includes both local and foreign company listings. In 2002, KSE won 'the best performing stock exchange of the world' award, while listing 592 companies with total capitalization of approximately 3518 billion Pakistani rupees (equivalent to $39 billion) (http://ksestocks.com/AboutKSE). Increasing foreign investment was seen in Karachi Stock Exchange in years 2006 and 2007 as it is an emerging market. According to the State Bank of Pakistan foreign investment in KSE increased by $523 million in year 2007 (http://SBP.gov.uk). However, the investment did not continued to increase because of worsening law and order situation and terrorism threats in the country which caused overall economic decline and financial tension. The increasing shortage of electricity and power sources in the country is another reason of downturn of industry in the country. However, the stock exchange is one of the important stock exchanges in the world and represents Pakistan in international financial world. It is important to conduct research on it and test effects of dividend announcement on stock prices. Weak form of market efficiency of Karachi Stock Exchange has been investigated by Ali & Akbar (2009), Husain (1999) and Chakraborty (2006) and they concluded that public information does not plays important role in determination of stock returns. Cash dividends are taxable in Pakistan while capital gains on equity markets are exempt from taxes. Therefore investigation of response of stock returns to different dividend announcements becomes important and relevant to stock markets in Pakistan.

In this research study event study methodology by Dolley (1933) and regression analysis by Legendre (1805) has been employed. Event study uses daily stock returns and an estimation window which is used to predict expected returns for event investigation window. Market risk adjusted model is used as a prediction tool. Effect of dividend announcement on share prices has been investigated for which prices and data has been obtained from Thomson One Banker, Karachi Stock exchange website and Data stream. For the period of five years from 2007 to 2011 total 40 dividend announcements have been selected based on pre-determined criteria including that the company did not announce earnings 90 days before or 20 days after the date of dividend announcement and company announced only annual cash dividends. To supplement the results derived from event study and check if other factors affect the stock price, regression analysis has been conducted of cumulative abnormal returns upon dividend increase, dividend decrease, payout, price to book ratio, return on assets and debt ratio. The study is significant as it will fill in the gap on literature for relevance of dividend announcements in Karachi Stock Exchange. There has been extensive research of this topic on stock exchanges of England, America, Greece, Turkey, China, Singapore, India and Bangladesh but not a lot of research has been done on stock exchange of Pakistan.

The rest of the research thesis is arranged in the following order. Part 2 covers existing research and literature review on the similar topic. Part 3 explains the data and how it has been collected for the purpose of analysis. It also describes the event study and its empirical results on whether dividend announcement has any effect on stock prices. Part 4 describes regression analysis and its results. Part 5 combine literature review and results from both research methodologies and then summarizes the results along with limitations of this research methodologies.

Literature Review

Much research has been conducted on impact of dividend announcements on the stocks returns. The results are mixed and ranges from no impact of dividend announcements on stock returns to positive and negative impacts when dividends increase or decrease respectively. This concept is also known as efficient market hypothesis which states that markets if are efficient will reflect all the information immediately, hence stock returns reflects dividends announcement as soon as they are announced. One of the earliest studies carried out in this regard are by Petit (1972) who found that market used dividend information in pricing securities. Whereas the phenomena exists from Lintner (1956), who states that managers believe that dividend changes are highly dependent on permanent earning changes and that managers have more information about the future cash flows and future earnings capacity of the company. Most important factor that managers consider on deciding dividends is company's future earnings (Leventis 2011). He also adds that managers wants to keep dividends stable and hence gradually changes to reach their targets. Therefore, the dividend announcement carries a lot of information for investors.

Researchers like Stevens (1989), Ariff and Finn (1986), Kato (1995), Gordon (1959), Lee (1995) and Ogden (1994) has found that stock markets show significant higher than average returns following dividend announcements. Frank (2004) conducted a study using 483 companies under regression analysis method and established that earnings ratio and dividend ratio have a positive relationship, which means that one of the factors effecting dividend payments is earnings of the company. The same author further established that there is positive relation between the debt to equity ratio and the dividend payments, which is the evidence of that debt to equity ratio is another factor effecting dividend policies of the companies. Paskelian (2006) adopted survey technique in explaining behavior of managers and concluded that stability of earnings, current earnings, future earnings, liquidity and financial leverage are prime factors that corporate managers consider while deciding on dividend payout.

Rozeff and Kinney (1976) explains that above normal returns are observed in month of January and companies release more information in January hence above normal returns can be attributed to this increased release of information in the markets. Lonie (1996) investigated dividend announcements of 620 firms in UK from January 1991 to June 1991 by means of event studies and found that investors responded to change in dividend whether dividends are increasing or decreasing. Fuller and Goldstein (2010) found that dividends matter more to shareholders in declining markets. Ball and Kothari (1991) studied dividend announcements and stock prices in US from 1980 to 1988 and found that abnormal returns follows dividend increase announcements. Foster and Vickery (1978) and Gordon (1962) confirm this view and documents evidence of positive abnormal returns after dividend declaration announcements. On contrary Easton and Sinclair (1989) have found statistically significant negative returns followed up by cash dividend announcements. This can be attributed to the income tax effect as cash dividends are taxed and tax on capitals gains can be deferred or is lower so investors put negative value on cash dividend announcements, hence price of stocks fall. Yunnan (2011) verifies it that decrease in taxes in 2005 in China led to increase in dividends by companies.

Acker (1999) investigates effect of dividend announcement on stock volatilities instead of stock returns and finds that stock volatilities around dividend announcement date increases particularly on final dividend announcement date. DeAngelo and DeAngelo (1990, 2006) studied dividend adjustments by troubled or loss making companies in New York Stock Exchange. They proclaim that there is a strong reluctance to omit dividends while companies only cut their dividends in periods of financial distress. Dividend cuts are observed after persistent losses while companies avoid cutting their dividends when losses are temporary or transitory. From this interaction of losses and dividend changes it is implied that some private information about manager's perception about the company is conveyed to outsiders through dividend announcements.

Uddin and Chowdhry (2003) investigated impact of dividend announcements on stock returns on Dhaka Stock exchange in Bangladesh using 137 companies announcing dividends from October 2001 to September 2002. The results for abnormal returns were statistically insignificant concluding that the dividend announcements had no information content in Dhaka Stock exchange. Ahmed (2010) adopted regression methodology to study effect of dividends on stock prices. His studies show that dividend yield and dividend payout have significant effect on the stock prices while leverage, size has negative but insignificant effect.

## Theories explaining relationship of stock returns to dividend changes

Many theories have been proposed to explain positive (negative) relationship of stock returns to dividend increase (decrease). Three of these theories are widely accepted: information signaling, free cash flow and clientele effect.

## Information Signaling

Since managers have more information about the future financial performance of the company, the announcement of dividends is taken as providing some information to the investors. Lintner (1956) suggests that managers will increase dividends when they are positively confident about the future performance of the company. However, it may also be taken as negative information that company does not have enough positive value adding opportunities to invest hence it is distributing cash to shareholders. It may also be taken as negative if shareholders are taxed higher on dividends compared to capital gains. In other words dividend announcement conveys important and valuable information about permanent changes in the earnings of the company.

John and Williams (1987) formalized this theory as "signaling theory" and suggests that changes in dividend reflect changes in expectations about future earnings of the company. Bhattacharya (1979) supports this by mentioning that dividends give information about the future cash flows of the company in imperfect market conditions. The company increasing dividend is claiming that it will have sufficient cash flows in future to sustain these dividends.

Watts (1973) tried to find out relationship between unexpected dividend changes and future earnings. Future earnings were forecasted using current dividends rather than earnings. Hi did not find any significant relationship between unexpected dividends and future earnings. Penman (1984) and Genodes (1978) also made the same conclusion in their studies. Bernartzi (1997) also studied the same and concluded that dividend policy is related to past earnings rather than future earnings.

Besides studies testing relationship between dividends and earnings it has also been investigated whether the dividend announcement has information content. Watts (1973) in his study, using monthly data rather than daily data, did not find any abnormal returns following dividend announcements. Modgliani and Miller (1961) suggested that in no tax world dividends does not has any effect stock returns as shareholders themselves can create dividends for themselves by selling or buying shares. However, Bernartzi (1997) in his study found evidence that increase in dividends cause positive abnormal returns while decrease in dividends cause negative abnormal returns.

Sometimes it is a problem to study effects of dividend announcements because earning announcements are made in close to dividend announcements. Swary (1980) developed research by Watts (1973) by overcoming its limitations and separating effect of dividend announcements from earnings announcements. He established that dividend does effect stock prices and has its own information content. The studies of Laub (1976), Woolridge (1982), Mullins (1983) and Travlos (2001) also confirmed that dividend announcement does carries information and effects the stock returns.

## Free Cash Flow Hypothesis

Another theory explaining relationship between dividend change and abnormal returns is free cash flow hypothesis arising from agency theory (Jensen 1976). Jensen and Meckling (1976) describe agency theory as separation of management from owners and each agent works in their own interest. Agency theory exists because of information asymmetry and conflict of interest among the management (agent) and the shareholders (principal). The free cash flow hypothesis by Jensen (1986) argues that management wants more free cash flow under their discretion to avoid the risk of bankruptcy, therefore are not willing to pay cash as dividends. Management may also have more perks and invest this cash in projects which are not actually value adding or are negative value projects. According to Eisenhardt (1989) problem here is that shareholders cannot verify if management has behaved appropriately in shareholder's interest who are actual owners of the company.

Jensen (1986) states that shareholders use dividends to monitor and discipline the management's action rather than direct intervention to management decisions. Therefore the increase in dividends is considered as positive information and reducing agency costs as management will have less cash available and it becomes less likely to invest in negative value adding projects. Similarly the increase in the dividends implicitly states that the future earnings of the company will be better in future. Thus the dividend announcement has an information content which aligns management's interests with those of shareholders of the company.

Easterbrook (1984) also supports free cash flow hypothesis. According to Easterbrook (1984) and Rozeff (1982) paying out dividends does reduce agency problems. High dividend paying firms often require external funding from the market. Therefore these firms are subject to stricter monitoring and evaluation by the investors in the market. It is often argued that free cash flow hypothesis is similar to dividend signaling theory and both convey information about the company to the market. However, the free cash flow hypothesis claims that information conveyed is related to behavior of the management rather than future expected financial performance of the company. Some studies have used Tobin-Q ratio to measure the overinvestment which distinguishes between dividend signal theory and free cash flow hypothesis. Litzenberger (1989) have used the same approach and documented in his study that free cash flows hypothesis plays better role than dividend signaling theory in explanation of relationship between dividends and its effects on stock returns.

## Client Effect Hypothesis

Third explanation for the relationship between the dividend changes and abnormal stock returns is client effect hypothesis. Some investors would like to have earnings paid out to them as cash so that they can use it or invest it on their own while others will want the company to retain the earnings so that they can have a capital gain together. Therefore some companies focus on shareholders who prefer earnings to be paid out and pay regular dividends, whereas some companies focus on shareholders who prefer earnings to be retained in the company and be reinvested. These preferences differ mostly because of tax regulations and different treatment of dividend income and capital gains income.

In many countries dividend income is taxed at higher rate than capital gains. It was the situation in United Kingdom until 1997 and in United States until 1986, but is no more effective. Similarly, in Pakistan capital gains are exempt from tax while dividend income is taxable. When the tax treatment on capital gains and dividend income is same then shareholders are indifferent whether the earnings are paid out as dividends or earnings are retained and reinvested within the company. Black and Scholes (1974) and Miller (1982) argued that there should not be any relation between the dividend changes and abnormal returns even if tax treatment is different on both capital gains and dividend income. Similarly Black (1976) in his paper "the dividend puzzle" claims that it is a puzzle that dividends have no effect on firm value but still companies pay dividends.

The investors, who are liable to pay higher tax on cash dividends, will prefer earnings of the firm to be retained rather than being paid out. In such a case, the announcement of the dividends will be taken as negative information because these investors will have to pay higher taxes in future, hence their response will be to short such shares and long non-dividend paying shares. Miller and Modgliani (1961) and Black and Scholes (1974) formulized this tax preference as tax clientele effect.

However, some investors may still prefer dividend payments rather than capital gains due to their personal preferences or tax exemptions for themselves. Some investors believe in a bird in hand is worth two in a bush which means that they prefer certain dividend over uncertain capital gains in the future. Pension funds and institutional investors need stable income to pay income and pensions to its holders since it is not easy to liquidate shares quickly. Verma (1994) and Short (2002) in their studies found positive relationship between the dividends being paid out and institutional ownership. Therefore the decrease in dividends is seen as a negative event by the institutional investors.

Some research has been conducted to test the effect of the dividend announcement on the value of the firm using dividend model by Brennan (1970). Litzenberber and Ramaswamy (1982) in their research concluded that dividend does have an effect on value of the firm while Black and Scholes (1974) found no evidence of dividends having any effect on value of the firm. Bajaj and Vijh (2002) argued that it is difficult to distinguish between dividend signaling and client effect hypothesis. Bajaj and Vijh (1990) also concluded that significance of effect of dividend changes on stock prices depends on the level of dividends previously being paid by the company. If a firm is paying high dividends and increases its dividends then it will have a more significant effect on stock price compared to if it was paying lower dividends. Dennis (1994) studied clientele effect and free cash flow hypothesis and confirmed that dividend changes does have an effect on stock prices.

Similar studies by Waymire (1994), John Lang (1991), Park (1995) and Cheng (2005) were also conducted to study insider trading. Such studies assume that dividend announcements do have an effect on stock prices. To test for insider trading the event study is carried out around dividend announcement dates. If share prices are found to be responding before the official dividend announcement date then it can be concluded that insider trading does exists and information is reflected in stock prices even before the announcement of the event.

## Existing research in Pakistan

A number of studies have been done on effect of dividend announcements on stock returns of the Pakistani companies. For example Nishat and Irfan (2001) in their study rejected the dividend irrelevance theory concept by Modgliani and Miller. They used 160 companies listed in the stock exchange during the period 1981 - 2000 and concluded that dividend announcement does have an effect on the stock returns of Pakistani companies listed in Karachi Stock Exchange. The researchers used dividend yield and dividend payout as a proxy for dividend payout and included these both factors along with earnings volatility, long term debt, growth and size in their regression model which attempted to explain price volatility. The researchers concluded that dividend yield does have negative impact on share price volatility. Similar results were found when joint earnings and dividend announcement effect was studied by Nishat (1992) for the period 1980 - 1986. He established that both dividend announcement and retained earnings have influential effect but dividend announcement effect is stronger than earnings announcement effect.

Kanwer (2002) also studied dividend policy of 317 firms listed in the Karachi stock exchange over 1992 - 1998 time period and used regression model with dividend yield as dependent variable and a dummy variable to proxy for signaling effect based on if earnings increased or decreased in the future. The author supported the signaling theory that future earnings have a positive relationship with the increased dividend yield.

Kaleem and Salahuddin (2006) conducted an event study to test the impact of dividend announcement on share prices in Lahore stock exchange using 24 firms over the period 2002 - 2003. They supported the Modgliani and Miller's dividend irrelevance theory that dividend announcement do not have significant impact on share prices. Mubarik (2008) studied 32 announcements in oil and gas industry over the period 2004 - 2008 and concluded same that dividend announcement does not have any effect on stock returns. However, Zaman (2007) analyzed effect of different events on share prices using multiple regression models but studying only prominent 6 firms listed in three stock exchanges of Pakistan. He concluded that announcement does have significant effect on share returns. Khan (2011) adopted regression analysis approach and studied effect of dividend payments on stock prices using 55 companies listed in Karachi stock exchange. His results were that return on equity, earning per share, dividend yield and profit after tax are positively related to share prices.

Different authors have reached different conclusions under different set of assumptions. Some accepted dividend irrelevant theory while many are proponent of dividend announcements having an effect on stock prices. The differences are due to different set of studies or small samples being taken in most of the studies.

Table 1: Showing existing research on dividend signaling in Pakistan

Author

Sample

No of Observations

Method of research

Findings / Results

Type

Time Period

Akbar and Baig (2010

79 companies listed in Karachi Stock Exchange

2004-2007

129

Abnormal returns using event study

*The stock returns were mostly negative

during the event window

*The abnormal returns were significantly positive during the event window of dividend announcements.

Mubarik (2008)

5 Companies in Oil and Gas Marketing sector

2004-2008

32

Event Study methodology

*Established insignificant negative values for AAR and significant negative value for Cumulative abnormal returns on the date of dividend announcement

*Also established that dividend and stock price had weak and inverse relationship among themselves

Zaman (2007)

6 companies listed in Karachi Stock Exchange

2000-2005

7

Event Study and Regression

*Stated that dividend announcement and earnings announcement has significant positive effect on share prices

Kaleem (2006)

24 companies listed in Lahore Stock Exchange

2002-2003

200

Event study using abnormal and cumulative abnormal returns

*Results were that dividend announcement did not had any effect on share prices and followed MMs dividend irrelevance theory

Kanwar (2002)

317 companies listed in Karachi Stock Exchange

1992-1998

2219

Regression using dividend yield and dividend changes as dependent variable

*Supported the signaling theory that future earnings are associated with the dividend yields

Nishat (2001)

160 firms listed on Karachi Stock Exchange

1981-2001

3200

Regression Model analysis using dividend yield and dividend payout as dependent variable

*Concluded that dividend policies does effect the share prices in Pakistan

Research

The study investigates the impact of dividend announcement by studying effect on stock prices followed by the dividend announcements using event study methodology. Event study has been employed several times to study the effect of an economic event such as mergers, earnings announcement and dividend announcements. In most of the studies, the focus is the effect of the event on the price of the particular type of share. Warner (1985) describes event study as an empirical research methodology to evaluate the effect of a particular event on stock price. In this study dividend announcement is an event and its effect on share prices will be investigated.

## Hypothesis

It needs to be tested whether the dividend announcement has significant effect on stock returns. This study will consider abnormal returns that are returns that the stock has earned in excess of returns that would have been earned if the event has not occurred. Hence, if these excess returns or abnormal returns are significant then it can be concluded that the dividend announcement does has an effect on stock returns. Contrary to that if abnormal returns are not significant then it can be concluded that stock returns are same whether the event happens or not therefore the dividend announcement does not have any effect on stock returns. Therefore the null hypothesis being studied in this paper is that abnormal returns are zero or in other words dividend announcement has no significant impact on stock returns. If null hypothesis is rejected then it will be concluded that abnormal returns are not equal to zero hence dividend announcement does have significant impact on stock returns. Formally null and alternate hypothesis will be stated as follows.

H0: Dividend announcement has no significant impact on stock returns or

abnormal returns are zero

H1: Dividend announcement does have a significant impact on stock returns

or abnormal returns are not equal to zero

## Data

Data has been collected from Thomson One Banker, Bloomberg and Data Stream. Companies with dividend announcement have been selected using the following criteria:

The company is listed and is not a financial investments firm because financial firms have many other factors effecting the stock prices hence effect of dividend announcement cannot be separated

Dividends announced are purely annual cash dividends and no other announcement has been made at the same time so that effect of only dividend announcement can be studied and other factors can be excluded

From the total of 569 companies listed in Karachi Stock exchange 40 announcements were selected (Appendix A) to match with above criteria. These 40 announcement event dates were then used to collect further data as described below in next part.

## Research Methodology

## Event Study history

An event study measures the impact of an event on value of a firm or prices of the share in the market. Event study is important and popular because it enables a researcher to test for significance of an event on underlying factor which is usually the stock returns or firm value. The event study methodology has many applications. In accounting and finance it has been applied to many firm specific and economy wide events. For example dividend announcements, earning announcements, mergers and acquisitions, share splits, issue of new shares or debts and macroeconomic events such as trade surplus, international agreements. Event studies have also been applied to many other subjects for example it has been used in field of law to measure the impact of new regulations on value of firm or to assess damages in legal liability cases (Schwert 1981). Focus of most of the studies has been to evaluate the effect of an event on the particular type of equity of the company such as common shares. Event studies were first used by Dolley (1933) in her study of effect of stock splits on stock prices. He used 95 splits and concluded that the effects were mixed but mostly the prices increased. Since first use of event studies in 1933 the level of sophistication of event studies has increased until 1960's when it was further developed by Brown (1968) and Fama (1970). Since then a number of modifications has been made which relates to complications arising due to statistical assumptions and accommodate more specific hypothesis tests. Two most important changes that have been made are the use of daily rather than weekly or monthly stock returns to improve efficiency of the tests and methods used to calculate abnormal returns have become more refined (MacKinlay 1997).

## Event Study outline

MacKinlay (1997) in his article 'Event studies in economics and finance' have mentioned event study methodology having basically seven steps. The first step is to define the event of interest and selecting the time period in which these events should be looked for. In this study event of interest is the dividend announcement and time period these events should be looked for is year 2007 to 2011. Second step is to establish criteria to select the companies involved in such events. These are defined under the heading Data below.

Third step involves calculation of normal and abnormal returns. Abnormal return is the actual return of the share minus the normal expected return over the event estimation window. Normal return is defined as the return the company would have earned if no event would have occurred. Different methods exists for computing normal returns and each of them has been discussed under the heading Abnormal Returns below. Fifth step involves setting up a framework to test for significance for abnormal returns such as conducting t statistic or z statistic tests. Sixth and seventh steps are presentation of empirical results and interpreting those results respectively.

## Event Window

Purpose of event studies is to test the effect of dividend announcements on stock prices. For each event an event window is defined. Event window is the number of trading days preceding and following the date of event i.e. Dividend announcement date. Researchers have not yet agreed on number of days to be included in the event window but lesser is better (Seiler 2004). For this study dividend declaration date is said to be t=0, and event window is defined of 40 days as t = -20 to t= 20.

To estimate the normal expected returns estimation window is required which is the time period when event has not occurred. It is used to establish returns that would have existed if the event would not have occurred. For this study the estimation period is taken as 90 days or t= -110 because three months before the event are considered appropriate to estimate stock prices for up to 20 days after the event and are not too far back so that they lose their prediction power. It is suitable because in Pakistan usually companies announce earnings or make other announcements quarterly hence to isolate the effect of dividends announcements from other announcements of different quarters it is important not to go further back than 90 days for estimation purposes. In literature many different event window and estimation periods has been used for example McCluskey and Burton (2006) used 180 days for Ireland, Naimat (2011) used 100 days for Pakistan while Padmavathy (2012) used 60 days for India. Following diagram represents event and estimation windows used for this study.

## Share Returns

For the selected announcements closing prices around the dividend announcement dates were collected from Thomson One Banker and Data Stream. These prices were then converted to annual returns using logarithmic formula instead of discrete formula. Strong (1992) mentions "...theoretically, logarithmic returns are analytically more tractable when linking sub-period returns to form returns over longer intervalsâ€¦and empirically, logarithmic returns are more likely to be normally distributed and so conform to the assumptions of standard statistical techniques". Following formula is used to calculate returns for the stocks and returns for the market.

Moreover, daily closing prices of stocks are used instead of weekly or monthly because daily data allows the impact of dividends to be isolated from the effects of other events (Brown and Warner 1985). However, some problems arise when using daily share returns instead of weekly or monthly share returns. Bown and Warner (1985) mentioned that daily returns cause statistical problems such as non-normality and autocorrelation as compared to weekly or monthly counterparts. It may also have higher risk of inconsistency and bias if daily data set is used rather than weekly or monthly. Therefore to minimize the problems associated with using daily data; logarithmic function is used to compute share returns which will minimize effects of non-normality.

Table2: Example for Arif Habib Corporation Limited

## Day

## Price

## Market KSE Index

-110

24.77

10533.57

-109

24.3

10586.46

Return on stock = ln =-0.01916

Return on Market index = ln = 0.005009

## Abnormal Returns

After these returns abnormal returns needs to be calculated. Abnormal returns are simply a difference between the actual returns and the normal returns, where normal returns are defined as the returns that would have existed if the event would not have occurred. Since, normal returns are not available as event as already occurred hence they need to be predicted. Several methods are suggested to predict normal returns; Mean return model, market return model, Fama French three factor model, index model and risk adjusted return model are some of these. However, all of them in essence suggest subtracting the predicted expected performance from the actual performance to calculate abnormal returns. What differs among these models is the assumption about the abnormal return and risk coefficients of the security.

Mean return model assumes that the mean of the stock's return over the event window is same as the mean of the stock's return over the estimation window. Under this model abnormal returns are defined as difference between the actual returns and the expected returns, where expected returns are mean returns over the estimation window. This model is very simple to use but fails to work when there is event clustering or market is rising or falling. It will cause estimates to biased in case of market is rising or falling (Seiler 2004).

Similarly, market return model assumes that the mean of the stock's return over the event window is same as the mean of market's return over the event window. Under this model abnormal returns are defined as difference between the actual returns and the expected returns, where expected returns are the mean returns of the market over the event window. This model does not require any estimation window and is suitable to be applied where there are no previous prices available such as in case of IPOs. This model is simpler and less hectic but may produce bias results in case of event clustering. Event clustering is when more events overlap and effects of each cannot be separated. Proxy return model is same as the market return model except that instead of market mean return it uses the mean return from the industry index. Proxy return model assumes that industry index better controls for the risk of the sample firms. If one firm is from high-tech sector, it is riskier than other firms then for this firm using industry mean is better measure to account for high risk.

Risk Adjusted return model is the most commonly used approach to compute expected returns for the event window. Abnormal returns under this model are defined as difference between actual returns and the expected returns, where expected returns are returns predicted from regression of stock returns on market returns during the estimation window. Issue under this regression is whether to use excess returns or normal returns. Those in favor of excess returns argues that it is consistent with the concept of Capital Asset Pricing Model, however most studies still use normal returns.

Brown and Warner (1985) argues that using more complicated and sophisticated models does not adds any value because the variance of abnormal returns is not significantly reduced by using any of these models. For this study Risk adjusted return model also known as market adjusted model will be used. Strong (1992) states this is most commonly used model and Warner (1985) mentions that this model produces same results even if share is traded infrequently. This model also produces smaller variances of abnormal returns and smaller correlations among abnormal returns giving better statistical results (Strong 1992). Moreover, the risk adjusted model also automatically accounts for the size of the firm (Schwert 1983).

Abnormal Return OR Estimation period firm residuals = Actual Return - (Î± + Market Return Ã- Î²)

Where Î± is calculated using Intercept function in Microsoft Excel by running a regression of market index returns on stock returns and Î²s calculated using Slope function in Microsoft excel using same regression function.

Table 3: Example for predicting Abnormal returns for Arif Habib Corp using Regression function

## Day

Actual Stock Return

Market Return

Abnormal Return of Share

-110

-0.019157

0.005009

-0.01752784

Intercept = Î± = -0.000297

Slope = Î² = -0.265954

Abnormal Returnt=-110 = -0.019157 - (-0.000297) - 0.005009 (-.265954) = -0.01752784

## Methodology & Significance Tests of TSAR

To test for significant t statistic or Z statistic can be used. If the abnormal returns are not standardized then t statistic is suitable while in this study the abnormal returns have been standardized hence z statistic have been used. Z statistic also needs only one standard value to compare with while t statistic require different values for different sample sizes hence makes z statistic easier to apply and simpler to interpret.

Choice of tests

Patell (1976) proposed abnormal returns to be standardized through dividing the abnormal returns by standard deviation so that the effects of returns with large variances can be reduced. Z statistic proposed by Patell (1976) assumes cross-sectional independence for abnormal returns and returns to be normally distributed. Campbell and Waslley (1993) have reported that test suggested by Patell has led to rejection of null hypothesis in NASDAQ exchange too often due to non-normality of Nasdaq returns, particularly securities which are less liquid and lower priced. Cowan and Seargent (1996) has also concluded that test suggested by Patell is too stringent and causes excessive rejections of null hypothesis.

BMP (1991) suggested variance-change corrected version of the test proposed by Patell. BMP's test became more popular because it has been found to be more robust in regard to possible changes in volatility associated with the event. For example, BMP (1991) have claimed that their test is correctly specified in New York Stock Exchange Amex even if variance of stock returns increase on the event date. Overall BMP and Patell's tests have become popular because they offer better power properties for significance tests.

Standardization of abnormal returns

Abnormal returns are standardized using following formula suggested by Seiler (2004) in his book 'Performing financial studies: a methodological cookbook'.

SARjt=

Where SARjt = Standardized Abnormal Return for company j at time t

ARjt = Abnormal Return for company j at time t

= standard deviation of the Abnormal Returns for company j at time t

Standard deviation is square root of the variance of the abnormal returns of firm j at time t. Variance is given by following formula:

## =

Where

= variance of Abnormal returns of company j at time t

= Abnormal returns of company j at time t over the estimation period

= average of abnormal returns of company j over the estimation period

= number of observed trading day returns for company j over the

estimation period

= Return on market index (KSE) at time t over the event window

= Return on market index (KSE) at time t over the estimation period

= mean return on market index (KSE) over the estimation period

The formula takes the difference of the estimation period abnormal returns from the mean abnormal returns and squares it so that negative values do not set-off positive values and total effect can be seen. It is then multiplied by the difference of market index returns from the average market index returns taking into consideration number of observations. Standardized Abnormal Returns (Appendix B) are then summed for each day to represent Total Standardized abnormal returns (TSAR).

Significance test

These total standardized abnormal returns are then tested for significance using z-statistic formula as below:

Z - Statistict =

Where

Z - Statistict = Z statistics test value for each day in the event window

TSARt = Total Standardized Abnormal Return for each day in the event window

Dj = number of observed trading day returns for firm j over the estimation period

N = number of firms in the same

Z statistic calculated as above is then compared to Z score normal distribution values given in a statistical table and if the absolute value of z statistic is greater than the value in the z statistic table the null hypothesis is rejected. Rejection of null hypothesis means that abnormal returns are not equal to zero hence the values are significant. Alternate to Z statistic is to calculate P values to determine significance of the TSAR. P-values are calculated as shown below for each day using NORMDIST function in Microsoft Excel. The null hypothesis is rejected if the P values are less than Î± and Î± in this study is chosen to be 0.05. Value of alpha is selected based on the 95% confidence level which is equal to 100 Ã- (1 - Î±).

Table 4: Showing TSAR and Z statistic values to test significance

## T

## TOTAL SAR

## Z Stat

## P Value

## 95%Â

## Â

## T

## TOTAL SAR

## Z Stat

## P Value

Â 95%

-20

1.757

0.275

0.784

No

## Â

1

25.4659

3.9805

0.0001

Yes

-19

-5.520

-0.863

0.388

No

2

41.4489

6.4787

0.0000

Yes

-18

1.560

0.244

0.807

No

3

1.3788

0.2155

0.8294

No

-17

3.120

0.488

0.626

No

4

-3.0344

-0.4743

0.6353

No

-16

7.425

1.161

0.246

No

5

7.3274

1.1453

0.2521

No

-15

5.189

0.811

0.417

No

6

-2.9266

-0.4575

0.6473

No

-14

0.799

0.125

0.901

No

7

10.5033

1.6417

0.1006

No

-13

13.029

2.037

0.042

Yes

8

-4.7716

-0.7458

0.4558

No

-12

1.287

0.201

0.841

No

9

-4.2365

-0.6622

0.5078

No

-11

-5.966

-0.933

0.351

No

10

4.2577

0.6655

0.5057

No

-10

-10.868

-1.699

0.089

No

11

-16.5022

-2.5794

0.0099

No

-9

-15.750

-2.462

0.014

Yes

12

-1.6478

-0.2576

0.7967

No

-8

-17.110

-2.674

0.007

Yes

13

-1.7421

-0.2723

0.7854

No

-7

-1.282

-0.200

0.841

No

14

-15.1778

-2.3724

0.0177

No

-6

5.581

0.872

0.383

No

15

-12.4673

-1.9487

0.0513

No

-5

-1.748

-0.273

0.785

No

16

-13.1487

-2.0552

0.0399

Yes

-4

-18.959

-2.963

0.003

Yes

17

3.2993

0.5157

0.6061

No

-3

-2.211

-0.346

0.730

No

18

-9.0250

-1.4107

0.1583

No

-2

5.410

0.846

0.398

No

19

-12.8516

-2.0088

0.0446

Yes

-1

5.750

0.899

0.369

No

20

-9.5892

-1.4989

0.1339

No

0

6.667

1.042

0.297

No

## Â

## Â

## Â

## Â

## Â

## Â

Among the 40 company announcements 5 were in 2007, 9 were in 2008, 13 were in 2009, 11 were in 2010 and 2 were in 2011. Most of the companies in the stock exchange announced dividends in 2009 and 2010 to signal the recovery from the economic recession and convey positive future performance of the companies in future.

Figure 1 - Showing TSAR and Cumulative TSAR

From the table 4 and figure 1 above it can be seen that Total Standardized Abnormal Returns (TSAR) increase significantly on the day of announcement and especially on day 1 and day 2 after dividend announcement was made at t=0. After which it decreased and then again increased significantly on seventh day of the dividend announcement. It can be seen that the days -13, -9, -8, -4, 1, 2, 11, 14, 16 and 19 has significant z statistic values. Given the values are less than 0.05 at days t=1 and t=2 the null hypothesis can be rejected and alternate hypothesis states that abnormal returns are not equal to zero which will be accepted. Therefore it can be concluded that the dividend announcements in Karachi Stock Exchange does have a significant effect on stock returns with statistically 95% confidence level. The abnormal returns on t= -13, t= -9, t= -8, t= -4 can be taken as either leakage of information or investors expecting dividend increase to be announced hence the activity in stock markets begin to increase even before the actual announcement of dividends. However, the activity was not so strong which depicts the uncertainty among the investors about dividend increases -they were unsure of dividend increase hence activity only increased slightly. Similarly, the effect was insignificant after two days until day 11 after the announcement of dividends, when investors became confident and caused increase in prices through buying pressure. Volumes of trading were also observed around the dividend announcement date and volume in trade for most of the stocks increased during first and second days of dividend announcement and then decreased slightly.

## Cumulative Abnormal Returns

Cumulative abnormal returns are calculated to analyze leakage of information before the event and to verify the results stated above. It must be assumed that daily abnormal returns are independent of each other so that appropriate significance test can be applied. Cumulative abnormal returns are simply the sum of the total standardized abnormal returns and is calculated using formula

Cumulative TSART1,T2 =

Where

Cumulative TSART1,T2 = sum of TSAR calculated previously

TSARt = Total Standardized abnormal returns in the event window

T1 = earliest date in the event window (-20 in this study)

T2 = later date in the event window (ranges from -20 to 20 in this study)

## Methodology & Significance Tests of Cumulative TSAR

To test the significance of the Cumulative TSAR following Z-statistic is used

Zt =

Where

Zt = the cumulative TSAR Z-statistics for each day in the event window

N = number of firms in the sample (40 in this study)

= Standardized Abnormal Returns of company j for each day in the event window

T1 = earliest date in the event window (-20 I this case)

T2 = later date in the event window (ranges from -20 through 20)

Dj = number of observed trading day returns for company j over the estimation period

Table 5: Showing Cumulative TSAR and Z statistic values to test significance

## T

## Cumulative TSAR

## Z Stat

## P Value

## 95%

## Â

## T

## TOTAL SAR

## Z Stat

## P Value

## 95%

-20

1.757

0.275

0.784

No

1

3.6261

0.3817

0.7027

No

-19

-3.763

-0.757

0.449

No

2

45.0750

1.4691

0.1418

No

-18

-2.203

-0.199

0.842

No

3

46.4537

1.7235

0.0848

No

-17

0.917

0.079

0.937

No

4

43.4193

1.3573

0.1747

No

-16

8.342

0.583

0.560

No

5

50.7467

1.8046

0.0711

No

-15

13.531

0.793

0.428

No

6

47.8201

1.4385

0.1503

No

-14

14.330

0.847

0.397

No

7

58.3234

1.9793

0.0478

Yes

-13

27.359

1.594

0.111

No

8

53.5518

1.5544

0.1201

No

-12

28.646

1.493

0.136

No

9

49.3153

1.7676

0.0771

No

-11

22.680

1.285

0.199

No

10

53.5730

1.5040

0.1326

No

-10

11.812

0.557

0.578

No

11

37.0709

1.4530

0.1462

No

-9

-3.938

0.150

0.881

No

12

35.4231

0.9638

0.3351

No

-8

-21.048

-0.912

0.362

No

13

33.6810

1.2515

0.2107

No

-7

-22.331

-0.464

0.642

No

14

18.5032

0.4889

0.6249

No

-6

-16.749

-0.676

0.499

No

15

6.0359

0.3852

0.7001

No

-5

-18.497

-0.433

0.665

No

16

-7.1128

-0.1828

0.8550

No

-4

-37.456

-1.420

0.156

No

17

-3.8135

0.0584

0.9534

No

-3

-39.667

-1.214

0.225

No

18

-12.8385

-0.3213

0.7480

No

-2

-34.257

-1.228

0.219

No

19

-25.6901

-0.4221

0.6729

No

-1

-28.507

-0.742

0.458

No

20

-35.2793

-0.8612

0.3891

No

0

-21.840

-0.745

0.456

No

## Â

## Â

## Â

## Â

## Â

## Â

Again same interpretation of Z values and P values is used in this table as explained before under the heading Methodology and Significance Tests of TSAR. It can be inferred from the table 5 above that all values of cumulative standardized abnormal returns were insignificant at 95% confidence level, except for seventh day of the dividend announcement. If confidence level is reduced to 90% then third day CAAR becomes significant whereas all days before the dividend announcement remained insignificant. This establishes that there was no leakage of information prior to the dividend announcement date. Even after the announcement the information effect was not so significant until seventh day of the announcement.

For event window (-15, 15) day 3 and day 5 become significant while for event window (-10, 10) day one and onwards become significant but none of the days before the event announcement are significant in both the cases, hence it can be concluded that there was no leakage of information before the event announcement. Cumulative abnormal returns after dividend announcement day are positive and remain positive until 15th day from the announcement. As can be observed from figure 1, the cumulative abnormal returns were rising from day of announcement and begin to decline from seventh day and turned negative on 16th day from the dividend announcement.

## Regression analysis

## Regression Analysis history

Earliest form of regressions analysis which was published by Legendre (1805) was method of least squares, which was then further developed by Gauss (1809). They both applied regression analysis to the study of orbits of the body about the sun. Gauss later in 1821 published the enhanced version of regression analysis known as least squares method. The term regression was suggested by Galton who believed it to be used in biological field only. In his works, Fisher (1922) challenged some of the assumptions of the regression function and introduced a better and robust version of regression analysis. Electromagnetic desk calculators were used to calculate regressions until 1960s and sometimes it took 24 hours to have the results of one regression. Today it has been developed to the extent that results of multiple regressions can be obtained in matter of seconds with more assumptions and control variables included. It has been now a major part of most of the empirical researches.

Regression function now is defined as an equation depicting relationship between the dependent and independent variables. These independent variables are assumed to be factors that have an effect on the dependent variables. It allows researchers to verify if independent variable has a positive or negative relationship with dependent variable and whether it is significant. Regression model is also employed for predicting and forecasting values based on the existing relationship of dependent and independent variables. For proper analysis of regression function a lot of assumptions including that of normality, non heteroskadisticity and no auto correlation is required which are usually difficult to achieve when there are small number of observations.

## Regression analysis application

To study the interaction among the factors other than dividend changes, that may have an effect on stock returns, following cross sectional regression model using Ordinary Least Squares method is being engaged. This method was first proposed by Kane (1984) for companies listed in United States and then was implemented by Kane (1991) for companies listed in Australia. Both concluded that these factors are often significant and does have effect on stock returns and should be considered in dividend announcement studies. Therefore these factors are included in the regression function.

CARit = Î²0 + Î²1DIit + Î²2DDit + Î²3Payoutit + Î²4ROAit +Î²5PBit + Î²6ATit

+ Î²7Debtit + Ôit

Where CAR is cumulative abnormal returns for the companies, DI is dividend increase dummy variable which is equal to 1 if dividend has been increased from previous year and zero otherwise, DD is dummy variable for dividend decrease which is equal to 1 if dividend has been decreased from previous year and 0 otherwise, Payout is dividend payout ratio (dividend/earnings), ROA is return on assets at the time of announcement, PB is the price to book ratio of the company, DR is the debt ratio in the books of the company (total liabilities/equity), AT is total asset turnover of the company, Debt is the debt ratio of the company while Ô represents the error term or the noise. It consists of all the residuals that are there and are not captured by any of the factors included in the equation.

The regression function sample consists of 40 announcements (Appendix A) to test whether cumulative abnormal returns does get effected by any dividend changes while controlling for other variable which also effect stock returns and indirectly the cumulative abnormal returns

Expected relationships

Dividend increase and decrease has been taken to see the effect of dividend changes on the cumulative abnormal returns. DI increase should be positively related while DD should be negatively related to CAR if we are to hold assumption that stock returns increase with dividend announcements and decreased with dividend decrease announcements. PAYOUT measures how much is being paid to shareholders, if dividend is giving positive message in the market then payout should be positively related to CAR. Dividends are normally considered positive news until or unless dividends are taxed at higher rate than the capital gains. In Pakistan, dividends are not taxed at higher rate hence higher payout should be positively related to Cumulative abnormal returns. Similarly Return on assets, price to book ratio, asset turnover and debt ratio effects stock returns hence are controlled for by being included in the equation. Higher return on assets is a measure of profitability and leads to higher stock returns hence ROA should be positively related to CAR. Price-to-Book ratio measures the value of the company relative to its book value. The higher priced companies should be earning higher stock returns hence should have positive relationship with CAR. However, the price-to-book value may be interpreted in a different manner such as the high value companies that is high price to book ratio companies may be regarded as having more investment opportunities hence investors willing their dividends to be reinvested so stock returns might not necessarily increase with the increase in dividends therefore the relationship with CAR can either be negative or positive. AT measures the asset turnover which is a measure of efficiency of the co

### Request Removal

If you are the original writer of this essay and no longer wish to have the essay published on the UK Essays website then please click on the link below to request removal:

Request the removal of this essay