Dividend policy has dominant practical effects to the discussion of corporate finance, so empirical study based on specific market would be beneficial to the further investment decisions made by market participators. The main purpose of this dissertation focuses on testing the dividend payout policys influence to the market value of company, and aiming to discover the regular pattern of dividend effects to stock prices. Event study has been applied to obtain the index of abnormal returns for sample firms in London Stock Exchange; and typical methodologies used to analyze the results contain descriptive information, parametric T test and Non-Parametric U test. The final tests results illustrate that dividends contain significant information for investors; especially decreased news has more important signal effects to investors, and the information before announcement date is more valuable to identify abnormal returns. Moreover, even though dividend policy would affect the stock prices, U test has showed that abnormal returns are observable but not attainable. Therefore, the discovery of this examination would be a valuable standard to evaluate the influence of dividend news to stock prices, and assist investors to make prudent and efficient trading decisions. Finally, the limitations of these tests include the methods to classify dividend news, dividend announcement information, forms of dividends and sample size. The future researches should consider these factors under different market environment for accurate and meaningful results.
Dividend policy is a dominant financial issue for analysts and corporate managers. One important reason is that deciding on the amount of earnings to pay out as dividends reflects the profitability and forecasting prospects of corporate companies. A proper understanding of dividend policy is crucial for many financial economists and analysts. Particularly, a view of methods and amount dividends are paid are beneficial to the analysis in theories of asset pricing, capital structure, mergers and acquisitions, and capital budgeting (Allen & Michaely, 1995).
Some of investors strongly believe that dividend information should always delivering positive investment signals. Dividends represent gross flows from corporations to their shareholders, so owners could use dividends to allocate resources out of the control of managers. Desai, Foley and Hines (2001) explained that corporate managers with discretion to select dividend levels can also use dividends to send credible profitability signals to the capital market, so the popularity of dividend policy can be understood.
But on the other hand, payout policy also be affected by different factors like tax burden, market efficiency and investor preferences, and could worsen some situations, especially the interest conflicts between managers and shareholders or bondholders and shareholders. These realistic elements have been taken into account and changed views of a large group of people who have prudent attitude assessing the differences between intrinsic share values and market values of companies.
This paper will illustrate a comprehensive understanding of the decision of dividend policy and the market effects of payout policy. Following Brealey, Myers, and Allens (2002) opinion, payout controversy issue is regarded as one of the top ten important unsolved problems in finance reinforces (Bhattacharyya, 2007). In a survey of dividend policy, Allen and Michaely (1995) indicate that sufficient empirical and theoretical researches play an important role in the discussion of dividend policy before reaching a consensus.
In this particular work, there are four major Chapters. Chapterhas briefly explained the topic of dividend policy in a view of past study; Chapter would discuss the dividend theories from different groups of views along with empirical evidences as literature review; in following Chapter , the major stylized empirical facts about dividend policy will be provided by event study that comparing different industries data based on London Stock Exchange Market, which relies on the transparency and liquidity of real time price of listed company. A proper conclusion will be made in Chapter about the usefulness and limitation of event study to dividend policy analysis as the final part.
. Literature Review
Brealey, Myers, and Allen (2008) described that there is a conservative group that believes the market reacts positively to announcements of dividend increases and negatively to announcements of dividend decreases. However, a radical group held opposite view that a higher payout reduces value as the tax effects. And in the centre, those who maintained market efficiency claim that payout policy makes no difference.
Therefore, it is difficult to conclude that a companys value is affected or not by its dividend policy. Realistic facts stress that companies would not supply a large quantity of dividends unless they believed that this was good for investors and the future development of company. Some believe that in our flawed world, a high dividend payout enhance share price and value of company (Brealey, Myers, and Allen, 2008, p 455-464).
Does the dividend policy change stocks fundamental value? If the answer is yes, investors would like to know how it affects the share value. The literature review of dividend policy would be theoretically discussed in three steps: namely original argument of dividend irrelevance; tax effects; and Dividend Payment Theories with Empirical reviews.
2.1 Dividend Irrelevance: M&M Theory
2.1.1 Corporate Practice of Lintners Survey
The first empirical study of dividend policy was provided by Lintner (1956), who surveyed corporate managers to investigate that what are the main principals for them to arrive at the dividend policy. Lintner concluded that an existing dividend rate decides a bench mark for the management. Companies management usually displayed a strong reluctance to reduce dividends. He pointed out that managers usually have reasonably definitive target payout ratios. Garrett & Richard (2000) proposed a behavioural model to assure the reality of target dividend level.
The study shows that managers are so concerned about the change of dividend policy because four stylized facts: Decent dividend paid by the company every year shows the stability of the performance of firm; and rational investors may pursuit the target payout ratio which indicates the consistency of companys earnings; smoothing is a phrase to describe a stable payout policy that would not supply unexpected dividend news; and finally if the dividend got largely cut, people who are not always rational are easily averse to even small losses (Lintner, 1956).
Besides of that, the behavioural finance explains some predicable ways which investors may do under changes of dividend. Psychologists claim that people prefer stable dividend change but not suddenly change because they are too rely on a few recent events in predicting the future. At the same time, some groups of investors may be overconfident in their predictions and unwilling to change the original choice (Brealey, Myers, and Allen, 2008). These behaviours explain the existence of abnormal return when dividend announced in reality. And some companies take advantage of this psychological logic and release adverse dividend signals to obtain unreasonable returns.
The One-bird-in-the-hand theory supports the result of survey by Lintner. Because the change of dividend payment is closely related to the interests of managers, investors are unable to access the interior information of the company, stable dividend policy provide accurate and visible returns of an investment. The long-term sustainable investment and growth objectives determined the firms target payout ratios. (Pandey & Bhat, 2007); they also pointed out that the Lintners model become acceptable in many developed markets for Canada, Australia, UK and France. However, the opponents hold opinions that the cost of positive signalling would be taxed heavier than forecast, so the One-bird-in-the-hand theory still stays in argument.
2.1.2 Corporate Practice of M&M Theory
After the research by Lintner, financial theory offers unsatisfactory answer to what is the optimal level and frequency of dividends announcement and which forms of payout could be the most indicative policy (Barclay and Smith, 1988). While Lintner (1956) provided the stylistic description of dividends, Miller and Modigliani (1961), initially proposed dividend irrelevance theory under the assumption that capital markets are perfect which without taxes, transaction costs and other market imperfections.
Given a firms investment program, the dividend policy of the firm is irrelevant to the firm value, and the profitable net present value projects would generate more cash flow which eventually increases the market value of company (Miller and Modigliani, 1961).Therefore, the companys investment decision is independent from financing policy, and no matter who undertakes the project and how it is ultimately financed.
When the boarder of company wants to facilitate a higher dividend for attracting more investors, it could necessitate more sale of stock or occupy creditors fund to raise finances for the investment program, so Miller and Modigliani (1961) stressed that the value of company could not be altered by dividend payment since the genuine reason of payable cash has been transfer from the old stockholders to new investors, which can be achieved by appropriate purchases and sales of equity (Allen and Michaely, 1995).
The crucial assumption is that the future market value will remain unaffected by current dividends. Miller and Modigliani (1961) demonstrated that in a perfect market setting, shareholders would be indifferent between share repurchases and the payment of dividends, indicating that debt is desirable and dividends are irrelevant to a firm's value if all forms of personal income are equally taxed.
Black and Scholes (1974) emphasised that dividend policy would not affect the value of company. If the annual dividends can be changed and controlled by the demand, company could definitely manipulate share price by adjusting the payout policy. Miller and Scholes (1978) explained that the interest amount can be set off against the dividend income by borrowing an appropriate amount, and reducing the taxable income to zero.
Miller and Scholes (1978) argued that the increases in risk due to borrowing can be countered by investing a risk-free insurance contract which matches borrowed amount, where the amount accumulates at the risk-free rate. As Bhattacharyya (2007) assessed, the tax shield on the interest expense can be used to neutralize the tax incidence on the dividend income without incurring any additional risk due to increased borrowing by this way.
Miller and Modiglianis dividend-irrelevance argument is elegant and ideal, and successfully explain the healthy form of a profitable and competitive company. The assumptions underlying M&M theory have five fundamental factors: (1) No taxes burden; stock market price cannot be changed by a single buyer or sellers; (2) No information asymmetry which requires that all the information is attainable and costless to access; (3) Unchangeable contracts between managers and shareholders which means no agency problem; (4) No transaction costs and the value of stocks are all based on the fundamental value of securities that discounted from future cash flow; (5) The market is complete that trading opportunities are equal with sufficient information as general.
However these factors cannot be fulfilled as the less perfect markets would have limitations in information and various costs. The reality is that most corporate companies, the public investors, investment analysts and experts are relying on the forms and amount of dividend announcements, which are considered to provide accurate forecasting reports in this flawed world.
2.2 Tax Effect under Market Imperfections
2.2.1 Taxes Payment and Transaction Costs
Historically, cash and stock dividends have been the predominant form of payout; share repurchases were relatively unimportant until the mid-1980s. In America, there was a dramatic difference between the capital gains and cash dividends before 1986 (Brealey, Myers, and Allen, 2008). Individuals in high tax brackets receive large amounts in dividends and pay substantial amounts of taxes on these dividends; meanwhile, dividend irrelevance is challenged and tax clienteles theory is strengthened, which make stock repurchase more advantageous than cash dividend in practice (Barclay and Smith, 1988).
Allen and Michaely (1995) emphasis that tax affect behaviour as the transaction fees reduce the payback rate. Different taxes burden could change the stock price around the ex-dividend day as well as the investors option. Dividend payment is considered as the cheapest way to get stable income from long-term investment. But the transaction costs for buying and selling stocks must be a decent amount of finance along with taxes. If the risks of investment cannot be effectively hedged, and then transactions costs increase the expense of total payment, tax burden affects the returns of stocks to a larger extend.
Al-Malkawi (2007) states that because transaction cost reduce the return of stock investment, high-tax investors naturally prefer the appreciation of stocks but no cash dividends; at the same time, small investors who do not have sufficient fund to finance the stocks are unable to pay the high transaction cost to sell security. These factors have impacts on the practice of M&M theory.
2.2.2 Tax Clientele Theory
As the existence of different levels of tax burdens, investors may have different preferences which are defined as tax clientele theory. Since pension funds are untaxed, those who could benefit from low tax rate could pursuit high dividend payout stocks; high-tax investors like retail investors have preferred low-yield stocks. However, under the proposition of perfect capital market, when people are pursuing that different dividend payout policies designed to minimize tax obligations, tax minimization cannot explain much of observed dividend policies.
Brealey, Myers, and Allen (2008) illustrate that when the aggregate demand of dividend payment is as much as aggregate supply, there is no company would change its market value by altering dividend payment policy. But if there are unsatisfied investors who pursuit high dividend payout rate, companies could be able to boost its share price by paying higher dividends. Whenever the company have stable payout rate, the stocks of this firm will attract a certain group of investors, so consistent dividend policy is beneficial to the long-term devolvement of company by the perdition of companys earning ability.
While formulating their famous dividend irrelevance propositions, Miller and Modigliani have tried to change the dividend level, according to different clientele groups of shareholders for the firm (Bhattacharyya, 2007). DeAngelo and Masulis (1980) analysed the recognition of personal tax rates that vary over both investors and sources of income. It is difficult to deny that taxes are critical in making investment decisions. They also simultaneously consider optimal debt policy, which is desirable because dividends and interest are alternative means for disbursing internally-generated cash flow to investors. DeAngelo and Masulis finally stated that debt is irrelevant but dividends may be relevant to a firm's value (Lally, 2010).
Fung and Theobald (1984) extend this analysis to dividend imputation systems. However, they ignore the implications of dividends for the level of share issues with their associated issue costs. A variant of the clientele theory has also been advanced to emphasise that dividend payments are in response to demands from investors for dividends (Baker & Wurgler, 2004).
There is some empirical evidence that U.S investors choice of stocks is oriented by the tax burdens. Allen and Michaely (1995) reported that the percentage paid out is higher measured as a percentage of net income than two decades ago; it was usually between 40%-50% in 1980s and compared to 30%-40% in the 1970s. Because there is a substantial tax disadvantage to cash dividends compared to repurchases. Cash payout policy effects are not as positively obvious as before under consideration of taxes, especially whenever dividends are taxed more heavily than capital gains, so that companies should pay the lowest cash dividend they can get away with (Brealey, Myers, and Allen, 2008).
Meanwhile, taxes on dividends have to be paid immediately while taxes on capital gains can be deferred until shares are sold and the capital gains are realised, so the longer the shareholders wait, the less the present value of the capital gains tax liability (Brealey, Myers, and Allen, 2008). Under this situation, shareholders have ability to control the moment when to pay the capital gains tax and optimal forms of dividends.
2.3 Dividend Payment Theories
2.3.1 Signalling Models
Corporations typically pay out a significant percentage of their earnings as dividends annually, and signalling models explain that dividend announcement convey valuable information. Miller & Modigliani suggested that if management's expectations of future earnings affect their current dividend payout decisions, then growth in dividends will convey information to the market future earnings. (Bhattacharyya, 1997).
London Stock Exchange also witnessed a growth of dividend payout from 2007 to 2012. From Appendix 1 which described last five years dividends issuing of FTSE 100, there are 67 companies enjoying positive dividend growth whereas only 25 firms reduce their dividends to some extend; to be more specific, 11% of those dividends have grown in more than 20% while 40% top 100 companies grew 10% in dividends. Regarding unexpected positive changes in current permanent earnings, Garrett & Richard (2000) have established a model to prove that dividends conveying information and the tests of signalling are sensitive to the specification of the model.
A group of natural clientele prefer high-payout security because dividends convey useful information about companies which can be regarded as an indicator of a firms profitability. An increase of dividend often proves the positive sustainability in earnings of a company whereas decreased dividends reflect the negative expectation of managers about the future development of business (Brealey, Myers, and Allen, 2008). Therefore, dividend changes may indicate different information content for investors.
They also states that high-dividend stocks are regarded as spendable income whereas capital gains are additional to principal. Because cash dividend or stock dividend is attainable and accountable for investors, large group of analysts give priority to approving these kinds of public stocks. If a company does not have good profitable investment opportunity, investors may expect higher dividends not because these are valuable in themselves, but as they encourage a more prudent, value oriented investment policy when managers are running the whole company for shareholders.
Signalling models have been developed to explain positive abnormal returns following announcements by firms of an increase in dividends (Bernhardta, Douglasb & Robertsonc, 2003). The signalling effect could be illustrated whenever to assess the share values and explanations of the dividend policies of firms. Unlike the financial reports which could be controlled by the management of directors, steady dividend signals convey pure forecasting information to the market.
2.3.2 Informational Asymmetry and Agency Effect
By studying dividend payments as signals, the existence of information asymmetry fosters the importance of agency problem. Under imperfect monitoring and information flow between owners and managers, the dividend decision could be affected and violate the development of company; these problems have become a significant issue in modern corporate finance which involving a large group of players: managers, employees, investors and shareholders.
Even though Barclay and Smith (1988) indicate that the merits of stock repurchase has been eliminated since the tax rate of cash dividend and capital gain becoming identical after 1986 in America; the information asymmetry between managers and shareholders are worsen as the increased arsing costs when company announces stock repurchase news.
Since corporate control problems and capital market signalling carry similar information, the outside investors would be misleading by the twisted positive or negative messages under signalling model, which cause the existence of abnormal earnings. McNichols (1989) revealed that the stock prices contain more accurate information than management forecasts which would not be fully reflected before trading days, so investors should take into account more factors and try to obtain more messages before deciding, and not only thinking about promised payout rates.
Agency problem increases the cost of information spread inside a firm and the public market. Under the hypothesis of efficient market, information could be sufficiently reflected in the market price eventually, but sometimes the managers are even reluctant to share information to shareholders as they could not attain any benefits; or the owners of company do not want to reveal the true financial difficulty to public, so it leads to the disparity about investor preferences or over reaction of investors when new information released.
When a company has limited profitable investment opportunity, shareholders could not trust the managers to spend retained earnings by enough consideration. Professional managers may utilize this free cash flow to benefit themselves which have occupied the financial resources of shareholders. High payout policy is a worthy way to get rid of agency problem for many companies, and balance the interests between managers and shareholders. This method ensures the efficiency and profitability of running projects by managers and at the same time eliminates the expenses to monitor managers by shareholders.
2.4 Empirical Results and Empirical Review
Empirical observations provide typical evidences and facts in the discussion of dividend policy. The leading theories of dividend policy could be classified into three majorities: tax clientele theory, signalling model and agency theory. The empirical support of these theories has been discussed in worldwide stock markets, and some particular hypothesis develops according to the financial and economic situation in different countries.
Considering the tax clientele theory, Allen and Michaely (1995) proposed that dividends should be minimized from the ex-dividend date under the consideration of tax perspective, which tells us that the taxes affect investment behaviour. At the same time, they indicates that there is no perfect holding clientele, and trading benefits and trading volume would be altered which means that high income individuals may invest in high-yield dividend stocks as well. An empirical study has been made to prove the existence of clientele theory by Graham and Kumar (2006); and they find evidence to support that lightly taxed institutional investors have high preference in trading the high-yield stock.
For the signalling model, dividends are used as an ex-ante signal of future cash flow (Bhattacharya, 1979). Moreover, dividends provide information about earnings as a description of the sources and uses of funds identity (Miller & Rock, 1985). The signalling model has been tested empirically with important implication: Dividend changes should be followed by subsequent earnings changes in the same direction; unexpected changes in dividends should be followed by expectations of future earnings in the same logic of dividend changing; unexpected dividend changes should be accompanied by stock price changes in the same direction (Allen and Michaely, 1995).
After the economic recession in 2007, Bozos, Nikolopoulos and Ramgandhi (2011) make an empirical study on the information asymmetry between stable economic condition and economic adversity. They compare the significance of average abnormal returns for dividend announcement and earnings announcement from London Stock Exchange, and conclude that dividends provide more information than earnings when the economy is regression but less information contend in periods of economic growing.
Lastly, agency theory has become to a dominant reason that company chooses to announce high dividend annually. Bhattacharyya (1997) conducted a model to explore that for a given level of earnings, lower type managers would declare higher dividends while profitable managers may prefer relatively lower dividends. Agency problem has been emphasised in his work that dividend policy would have close relation to the presence of moral hazard.
Dividend behaviour of companies in regulated and emerging markets like China and India becomes to a popular and contentious issue among researchers. Agency theory is no longer just the relationship between managers and shareholders but also between the government and company themselves. Pandey & Bhat (2007) proposed a designed research and findings that dividend policies of Indian firms are affected by the restricted monetary policies. Their methodology in discussing the dividend policy in emerging market suggest that macroeconomic policy has an indicative impact on the financial decisions which could cause a 5%-6% reduction or increase in payout ratios.
2.5 Hypothesis based on London Stock Exchange Market
2.5.1 Why choose London Stock Exchange Market
The choice of London Stock Exchange Market enjoys considerable dimensions: (1) London Stock Exchange market is a developed stock market, which provides us more accurate and transparent statistics. (2) Britain had the best quality accounting information among western countries (Hannab, 2007); and companies have published much more well-regulated and better information than required for research. (3) There are much more empirical researches about the New York Stock Exchange market in dividend policy than London Stock Exchange Market no matter before the financial crisis in 2007 or after that. Therefore, this thesis will focus on the dividend policy effect under the macro economy environment of recovery. All the samples would be listed companies in the London Stock Exchange Market, and the annual reports in time spread last five years from 2007 to the very lately 2012.
The significance of this empirical test depends on the accurate daily stock prices and ex-dividend dates information which are downloaded from Database. Moreover, the index of FTSE 100 will be regarded as the average return of market from the London Stock Exchange website. The index of FTSE 100 comprises the most highly capitalised blue chip companies, which have represented about 81% of the value of the UK market. Some analysts reports from Thomson One Banker and the dividend announcement news could be found from Nexis.UK. The dividend change announcement is an independent report, and the announcement provides public information about the frequency and level of payout rates and distribution, so it is an accurate material for testing market's response to dividend policy.
As cash and stock dividend is the most popular dividend payout form, the hypothesis of empirical test would focus on the mixed form of dividends paid by five different industries in London Stock Exchange Market, namely Food & Drug Retailers, Real Estate, Banks, and Gas, Water & Multiutilities and Automobiles & Parts. From the graphs in Appendix 3, oil and gas industry ranks as the first greatest weight among 100 companies while the Banks take 12.09% thirdly. Besides of that, Automobiles & Parts and Real Estate are listed as an indispensible part of FTSE 100. Therefore, the sample industries of this research have significant representativeness of London Stock Market.
2.5.2 Hypothesis
To simplify the sample groups for testing the impact of earning announcement, the dividend events would focus on the latest dividend announcement activity which mainly happened on the first quarter of 2012 (see Appendix 2). The specific dividend announcements are selected because they are the newest dividend policy record for these companies, and which may obtain the most accurate result of this assumption.
In order to improve the analysis of dividend news effect to companys equity, the announcements have been categorized into three groups: positive increased news, no news and negative decreased news. From the views of investors, higher dividend than expected should be associated with higher abnormal earnings and lower dividend news would lead to lower earnings.
Therefore, the classification of samples could use the deviation between actual dividend earnings and expected dividend earnings regarding the dividend news. When the dividend released has been increased in 2.5% than expected is named positive increased news; and if the dividend announcement has 2.5% deviation lower than expected dividend will be classified as negative decreased news; lastly, the middle group of dividend announcement is regarded as no news (MacKinlay, 1997). The hypotheses have been listed as follows:
H1: The market value (stock price) of Company would significantly increase when firms make positive increased dividend announcement.
H2: The market value (stock price) of Company would decrease when firms make negative decreased or no dividend announcement.
H3: The market value (stock price) of Company would fluctuate significantly when new dividend announcements are made.
If dividend announcements convey useful information to market participators, the announcement of dividend will impact the market value of companys stock. That is to say, the relation between dividend information and abnormal return will be discovered if the value of equity has changed, especially the stock prices become much higher than expected when positive information released or the stock prices become much lower when negative information comes to public market.
Chapter Data and Methodology
3.1 Methodology
3.1.1 Introduction of Event Study
Event study is an ideal tool to reveal the financial policy effects in most reports. The objective of empirical test is to perform an event study for a set of UK firms that announced dividends very recently. Event study could prove the existence of significant stock price changes observable for UK announcing dividends, and whether the market values of companies changing or not as dividend announcements. That is to say, the usefulness of event study will be recognised as a popular choice which is able to provide rational evidence to highlight the truth.
Before applying event study, we should be cautious about the statistical assumptions which would lead to modified results. Due to the scale limit, the sample size is not as large as expected and the abnormal return will be affected by more other factors like industry characteristics, estimated period and the financial environment during that period.
In this research, event study would be applied to examine the stock price effects by dividend announcement. There are 54 list companies which are chosen as the sample of this study from five different industries, while only 39 companies are identified through the extensive research from Nexis.uk database (see Appendix 2). Regularly, these kind of news are normally took place semi-annually by large public firms whereas annually by smaller companies in London Stock Exchange; for instance, Sainsbury have announced dividends 10 times during the 2007/7/25 to 2012/7/25, which is the main period of this event study.
Table 1: The List of Sample Companies in Dividend Announcement Groups
Positive increased news Negative decreased news No news
SAINSBURY PLC BLD PROPERTY HLDGS ASEANA PROPERTIES
DAIRY FARM INTL.HDG. SOLIDERE UNITED DRUG
GREGGS TESCO ROYAL DUTCH SHELL
JARDINE MTSN.HDG. MORRISON (WM) SPMKTS. DEE VALLEY GROUP
OJSC MAGNIT GDR THORNTONS SSE
O'KEY GROUP GDR BP RENEWABLE ENERGY
APR ENERGY CENTRICA (~ ) GKN
NATIONAL GRID (~ ) DRAX GROUP (~ ) VOLKSWAGEN AG
PENNON GROUP SEVERN TRENT (~ ) HONG KONG LAND
UNITED UTILITIES LLOYDS BANKING GROUP TOYOTA MOTOR
HSBC BARCLAYS
ROYAL BANK OF SCTL.GP. HYUNDAI MOTOR
STANDARD CHARTERED HONDA MOTOR (~ )
AMTEK AUTO (~ ) JERSEY ELECTRICITY 'A'
TOTAL PRODUCE (ESM)
Total: 15 14 10
A total of 316 dividend announcements news was released by 39 sample firms during the exam period. We could obtain essential data from the released news by analysts during 2007 to 2012. Dividends are closely relative to the performance of individual company annually, and the final dividend could be proposed and approved by the board, so before the record date and payment date, ex-dividend date is regarded as the event date.
According to the classification of hypothesis in section 2.5.2, among the latest dividend announcement of 39 companies, there are 15 firms which have positive increased news, 14 with negative decreased news and remaining 10 no news (see Table 1).
3.1.2 Procedures of Event Study
The first step of designing the event study is to identify the event (announcement) date, which is given in the Appendix 2. Because the effect of these kinds of announcement will last longer than expected, MacKinlay (1997) described that the event windows should be defined as a multiple dates period, which should be 20 days earlier and later than the event date. Under this period, no matter before or after the event date, the abnormal earning could be investigated by discussing the returns from different periods.
In this study, day 0 is regarded as the announcement date and 20 trading dates before and after the announcement date are event window; furthermore, the estimation period which are used to obtain the parametric data will be 100 days prior to the event window, that is to say, day -120 to day -21 is the estimated window in this research (see Diagram 1).
Diagram 1: Description of Event Study
Pre-Event Period Event Date Post-Event Period
-4 -3 -2 -1 Event Date 0 1 2 3 4
Ratios: Mean (Median) Ratios: Mean (Median)
Generally speaking, the estimation period could not include the event window for preventing the influence to the estimation of market parameter model. If the estimation includes the event window to estimate the normal return parameters, the abnormal return would be influenced as the expected return have been impacted by dividend announcement event.
The following step of event study is to define the statistics which are suitable for further analysis. Since different industries have specific characteristics, the collections of companies information are valuable for analysing the abnormal returns. Moreover, detail investigation about the macro economy during those five years is significant for relating statistics and conclusion. Therefore, industry information, distribution of dividend announcement and even the information of broad have to be considered during the information definition (MacKinlay, 1997).
For each sample firm, we could calculate the actual daily returns by following formula:
Rjt=In{ (Pt+Dt)/Pt-1} (1)
Where Pt is the stock price on the present day during the estimation period and Pt-1 is the stock price on the previous day and the Dt is the dividend paid on ex-dividend day; and most trading days Dt would be 0. By the same logic, the daily return of the market index could be calculated by following formula:
Rmt=In (It/It-1) (2)
It is the closing value of the FTSE 100 on the date t and It-1 is the previous date of specific window.
When we use the LINEST function by defining Rmt as the X and Rjt as the Y, the estimation of parameter result comes for j and j, so the market model for getting parameter estimates comes as following:
Rj=j+jRmt+j (3)
Where Rj and Rmt are the daily returns of one stock and the daily return of market portfolio, and j should be 0 as disturbance term. This market model has advantages by removing the part of return that could be varied from the market return, which means the variance of abnormal return is reduced (MacKinlay, 1997). Therefore, the usage of market model will be benefiting from the Rj market model regression, which indicates that the higher Rj the greater variance reduction and the larger returns. By ordinary least squares (OLS) for calculating the market model parameters, the estimation data will theoretically correct for calculating the returns in event window.
The most important step of event study is to test whether the abnormal returns are significant from 0 during the trading days in event period. Abnormal returns are defined as difference between actual returns and estimated returns.
ARjt=Rjt-j-jRmt (4)
The null hypothesis, H0, would play an important role in the initial analysis of abnormal return. When ARjt equals to 0 in the event window, there is no difference between the estimation and reality. This method can be used to test any period during event window (MacKinlay, 1997).
Moreover, aggregate abnormal returns for individual company or through time series are important to discuss the dividend policy effects for different companies. At the beginning, it is necessary to add all the AR for a stock, and then aggregating the AR in the specific period during event window. The significance of this cumulative abnormal return (CAR) is that could be useful when analyzing any period of dividend announcement.
CARARjt (5)
The cross-sectional average abnormal return for each trading day in event window could be calculated as:
ARt = 1/NARjt j= 1, . . . , N t=-20,-19,, 20 (6)
This data as testing the null hypothesis could eliminate the errors when applying the null hypothesis. Under the application of a cross-sectional regression, Benartzi (1993) explain that dividend changes have some predictive power about subsequent quarterly earnings changes.
Parametric tests could be applied to prove the significance of daily abnormal return and cumulative abnormal return in mean for H1, H2 and H3. Specifically, SPSS is the principal software to run 1-sample mean T test under the hypothesis for the average AR and selected trading days CAR of sample groups.
Ho: Mean ARt=0 or Ha: Mean ARt 0;
H1: Mean CARt=0 or Hb: Mean CARt 0
The confidence interval of this test would be 99%, 95% and 90%, which represents significance at the 0.01, 0.05 and 0.1 level respectively, and marked in ***, **, * separately. The conclusion could be presented as Null: There is no significant difference between the sample mean and 0; alternate: There is a significant difference between the sample mean and 0.
Furthermore, Non-Parametric Tests are necessary to test changes in median for H1, H2 and H3 as well. Originally, Mann-Whitney U test (Wilcoxon signed-rank test) is one of most popular methods as non-parametric statistical hypothesis test. It is valuable to assess whether one of two samples tends to have higher values than the other one. So the Hypotheses are following:
H0: Median1 = Median2 H1: Median1 Median2
If the AR and CAR before and after the dividend day is same, the result would prove that there is no significant difference between real return and estimated return which dues to the dividend announcement.
Finally, the industry analysis based on these data would be remarkable for getting a proper conclusion. The understanding and further discussion about the dividend effect and related reasons are an important objective for implementing event study. If the particular dividend news has a correlation to the value of the firm or stock price, it is certain that the accounting information content means much more than expected.
3.2 Data Analysis
3.2.1 Result Presentation and Discussion
Event study is built to effectively discover the significance of abnormal return by analysing data. The objective of this research focuses on the influence of dividend announcement in the London Stock Exchange Market after the economic regression. Even though previous studies have announced that dividend policy have significant effects to the stock price in developed markets, the extent of this impact would not be as obvious as before because the investor preference and economic background have changed.
Table 2: AR & CAR for Three Grouped News by Market Model in Event Window
Event
day Positive increased news Negative decreased news No news
AR CAR AR CAR AR CAR
-20 0.0051 0.0051 -0.0038 -0.0038 0.0051 0.0051
-19 -0.0015 0.0036 0.0047 0.0009 0.0015 0.0066
-18 -0.0041 -0.0005 -0.0032 -0.0023 -0.0019 0.0047
-17 0.0010 0.0004 0.0030 0.0007 0.0010 0.0057
-16 -0.0008 -0.0003 -0.0033 -0.0026 -0.0062 -0.0005
-15 -0.0056 -0.0059 -0.0055 -0.0082 0.0003 -0.0002
-14 -0.0026 -0.0085 0.0041 -0.0041 0.0006 0.0004
-13 0.0071 -0.0014 0.0014 -0.0027 0.0006 0.0010
-12 0.0035 0.0021 -0.0013 -0.0040 -0.0009 0.0001
-11 -0.0039 -0.0019 0.0011 -0.0029 -0.0006 -0.0005
-10 -0.0020 -0.0038 0.0004 -0.0025 -0.0070 -0.0075
-9 0.0018 -0.0020 -0.0019 -0.0044 0.0105 0.0030
-8 0.0000 -0.0020 -0.0054 -0.0098 -0.0126 -0.0096
-7 -0.0045 -0.0065 -0.0060 -0.0157 0.0009 -0.0087
-6 -0.0035 -0.0100 -0.0010 -0.0168 0.0043 -0.0043
-5 -0.0052 -0.0152 -0.0020 -0.0188 -0.0147 -0.0190
-4 -0.0040 -0.0192 -0.0030 -0.0218 0.0043 -0.0147
-3 0.0033 -0.0159 0.0036 -0.0182 -0.0037 -0.0183
-2 -0.0033 -0.0193 -0.0066 -0.0248 -0.0041 -0.0224
-1 -0.0008 -0.0200 0.0146 -0.0102 -0.0043 -0.0267
0 0.0142 -0.0058 0.0108 0.0006 -0.0060 -0.0327
1 0.0008 -0.0050 0.0015 0.0021 -0.0062 -0.0389
2 -0.0050 -0.0100 -0.0068 -0.0047 0.0033 -0.0356
3 -0.0021 -0.0120 0.0010 -0.0038 -0.0054 -0.0410
4 0.0016 -0.0104 -0.0004 -0.0041 0.0030 -0.0381
5 0.0037 -0.0067 -0.0066 -0.0107 0.0027 -0.0353
6 0.0001 -0.0066 0.0012 -0.0095 -0.0074 -0.0427
7 -0.0040 -0.0106 -0.0075 -0.0170 -0.0046 -0.0473
8 -0.0063 -0.0169 -0.0071 -0.0240 -0.0058 -0.0531
9 0.0022 -0.0147 -0.0004 -0.0245 0.0091 -0.0440
10 0.0000 -0.0147 -0.0017 -0.0262 -0.0090 -0.0530
11 0.0047 -0.0100 0.0024 -0.0238 -0.0008 -0.0538
12 0.0015 -0.0086 -0.0016 -0.0254 0.0006 -0.0532
13 -0.0048 -0.0134 -0.0036 -0.0290 0.0007 -0.0525
14 0.0076 -0.0058 -0.0002 -0.0292 -0.0049 -0.0574
15 -0.0061 -0.0119 -0.0030 -0.0322 -0.0062 -0.0636
16 0.0019 -0.0100 0.0068 -0.0254 0.0054 -0.0582
17 -0.0023 -0.0123 0.0023 -0.0231 -0.0042 -0.0624
18 0.0032 -0.0091 -0.0058 -0.0289 -0.0005 -0.0628
19 0.0066 -0.0025 -0.0039 -0.0328 0.0026 -0.0602
20 0.0080 0.0056 -0.0025 -0.0353 -0.0084 -0.0686
Referring the event window, it is necessary to aggregate the daily return across the days and stocks. Table 2 above shows the aggregative return (AR) from -20 trading days to +20 trading days of sample companies; and CAR represents the aggregation of abnormal return for three news groups respectively.
The sample result presents the latest dividend announcement return in 2012 for the 39 companies. All the market parameter estimates of individual company are calculated separately by the standard LINEST regression model in estimation window from t=-120 to t=-21. Since the dividend announcements have three different groups: positive increased news, negative decreased news and no news; AR presented is the aggregation of grouped samples average abnormal return in the specific trading day while the CAR showed is the daily cumulative abnormal return from day -20 to day 20 for different classifications.
The statistics indicates that positive increased dividend news has the highest abnormal return among three groups, which is consistent with the hypothesis that dividend news contains information for investors. On the dividend announcement date (day 0), the average abnormal return is 0.014204 of 15 positive dividend news sample firms, and 0.010791 in 14 negative dividend news securities, lastly -0.00599 in no news. Moreover, only the first grouped dividend news has positive CAR in 0.005563 whereas the other groups witness a negative CAR below 0 (-0.03525; -0.06859) for these 40 trading days.
Diagram 2: AR for Earning Announcement from Day -20 to Day 20
From diagram 2 which describes the daily abnormal return in event window, more trend developments of these three categories are observable. With the fluctuation of daily return, the abnormal returns of positive increased news are almost ranging from -0.006 to 0.015; but the AR of no news are negative in 80% of the event days. An insignificant effect by negative news group has been found that positive AR is even higher than positive news group in day -13, -11, -1 1 and 6. To the end of event window, only the positive increased news group have positive AR (0.008011) on day 20; while the other groups share the opposite trend that AR are negative. These results are consistent with the hypothesis 1 and hypothesis 2 as well.
Diagram 3: CAR for Earning Announcement from Day -20 to Day 20
The lines in diagram 3 show that the reaction of market is significantly relative to different dividend news especially after the announcement date. Before day 0, the CAR of three dividend groups have similar trend with slight decrease; focusing on the period after announcement day 0, the cumulative abnormal returns have developed in totally different directions between day [-1, 20]. To be more specific, the positive increased news boost the CAR in day [1, 2] and keep growing to above 0 in day 20; on the other hand, negative decreased news have a short increase around trading days [-2, 2] but dropping down at the end of day 20. Therefore, market price for decreased dividend news has relatively accurate reaction especially after the information totally releasing to the market. The interesting point is that there is a large decrease of no news groups CAR which has fell even heavier than negative news, and ending lowest point in the graph. The interpretation of the negative reaction of market to the no news is that investors are expecting to obtain dividend news in regular pattern; whenever the omission of dividend occurs in seasonal or annual financial reports, these disappointing signals could be interpreted as bad news. That is the reason why no news leads to more serious drop of stock prices. In summary, increased dividends news have positive effects while the decreased dividends or no news lead to a drop of stock market value by the basic descriptive comparison and data analysis.
3.2.2 Measuring and Analysing by Parametric T-test.
Parametric test is a method to test statistics with single factor null assumption, and provides evidence for supporting that whether dividend events have impact to the return of stocks. If the T possibilities range from 0 to 0.1, there would be different significant levels in 99%, 95% and 90% confidence intervals for the dividend announcement effects, which further conclude that reject original hypothesis and accept Nonzero hypothesis.
3.2.2.1 Hypothesis 1: The market value (stock price) of Company would significantly increase when firms make positive increased dividend announcements.
Table 3 has showed the T-value and significance level for abnormal return and cumulative abnormal return regarding the positive dividend news. Test details could be found in Appendix 4. According to the data on the left hand side, there is no obvious abnormal return around the announcement date. Even though the T-value is positive in 0.859 on day 0, the significant level stated (0.405) is much higher than 0.1, implying that increased dividend news cannot affect stock price significantly on the event date, which is opposite to the discussion in 3.2.1.
Table 3: The Stock Market Reaction to Positive Increased Dividend Announcements (See Appendix 4)
Event
day Positive Increased News
AR CAR
T-value sig.(2-tailed) T-value sig.(2-tailed)
-20 1.606 0.131 1.939 0.071*
-19 -0.366 0.72 0.39 0.702
-18 -1.18 0.258 0.088 0.931
-17 0.328 0.748 0.11 0.914
-16 -0.183 0.857 -0.219 0.83
-15 -1.95 0.072* -1.248 0.231
-14 -0.804 0.435 -1.139 0.273
-13 2.399 0.031** -0.133 0.896
-12 0.734 0.475 0.284 0.78
-11 -0.767 0.456 -0.423 0.678
-10 -0.339 0.739 -0.217 0.831
-9 0.298 0.77 -0.046 0.964
-8 -0.003 0.997 -0.118 0.907
-7 -0.735 0.474 -0.413 0.685
-6 -0.616 0.548 -0.434 0.67
-5 -1.801 0.093* -0.796 0.438
-4 -0.95 0.358 -1.056 0.308
-3 0.706 0.492 -0.708 0.49
-2 -0.732 0.476 -1.089 0.293
-1 -0.295 0.773 -1.049 0.311
0 0.859 0.405 -0.282 0.782
1 0.18 0.859 -0.211 0.836
2 -1.044 0.314 -0.396 0.698
3 -1.015 0.327 -0.605 0.554
4 0.33 0.746 -0.467 0.647
5 0.937 0.364 -0.282 0.782
6 0.021 0.984 -0.319 0.754
7 -0.586 0.567 -0.45 0.659
8 -1.075 0.301 -0.594 0.561
9 0.408 0.689 -0.645 0.529
10 0.003 0.998 -0.549 0.591
11 0.632 0.538 -0.519 0.612
12 0.469 0.646 -0.456 0.655
13 -1.18 0.258 -0.631 0.538
14 1.069 0.303 -0.309 0.762
15 -0.805 0.434 -0.619 0.545
16 0.362 0.723 -0.648 0.527
17 -0.457 0.655 -0.581 0.57
18 1.059 0.307 -0.57 0.577
19 1.091 0.294 -0.06 0.953
20 1.503 0.155 0.205 0.84
***, **, * represents significance at the 0.01, 0.05 and 0.1 level respectively
Nevertheless, the figures before the announcement date have implied a relatively significant effect on day -15, -13, -5. Specifically, day -13 enjoy a 0.05 (0.031) significant difference level with 2.399 t-value; day -13 and day-5 have t-value of -1.95 and -1.801 respectively at the significance of 10%. Changing to the CAR T-test, only 20 days before announcement date has t-value of 1.939 in 0.1 significance level.
The statistics describe a complicated trend for the discussion of positive dividend effect. The abnormal earnings are not occurring after the dividend information have been published and understood by market, but existing in more than 10 days advance. Moreover, positive t-value comes in interval tempo but not consistent at all. These phenomena may happen because the market has not received confirmed dividend news ahead of time but positive information have been released to the public. After day 0, there are more positive t-values until day 20; however, the amount has not reached significant level. The reasons about this result may be considered by composite factors: investor confidence, macroeconomic environment and dividend forms in London Stock Market. It would be discussed further with the correctness of hypothesis 1 after all the tests.
3.2.2.2 Hypothesis 2: The market value (stock price) of Company would decrease when firms make negative decreased or no dividend announcements
Table 4: The Stock Market Reaction to Negative decreased Dividend or no news Announcements (See Appendix 5)
Event
day Negative decreased or no news
AR CAR
T-value sig.(2-tailed) T-value sig.(2-tailed)
-20 -0.02 0.984 -0.02 0.984
-19 0.824 0.418 0.74 0.467
-18 -0.952 0.351 0.13 0.897
-17 0.771 0.449 0.509 0.615
-16 -1.054 0.303 -0.219 0.828
-15 -0.713 0.483 -0.667 0.511
-14 0.619 0.542 -0.335 0.741
-13 0.19 0.851 -0.116 0.909
-12 -0.305 0.763 -0.206 0.839
-11 0.111 0.912 -0.181 0.858
-10 -0.596 0.557 -0.351 0.729
-9 0.838 0.411 -0.101 0.92
-8 -3.417 0.002*** -0.775 0.446
-7 -0.974 0.34** -1.065 0.298
-6 0.492 0.627 -0.967 0.343
-5 -2.057 0.051 -1.673 0.108
-4 0.019 0.985 -1.799 0.085*
-3 0.139 0.891 -1.918 0.068*
-2 -1.995 0.058* -2.483 0.021**
-1 1.111 0.278 -1.288 0.211
0 0.405 0.689 -0.836 0.412
1 -0.681 0.503 -0.885 0.386
2 -0.863 0.397 -1.017 0.32
3 -0.362 0.721 -0.997 0.329
4 0.225 0.824 -0.912 0.371
5 -0.675 0.506 -1.071 0.295
6 -0.818 0.421 -1.173 0.253
7 -1.978 0.06* -1.403 0.174
8 -1.915 0.068 -1.62 0.119
9 1.031 0.313 -1.485 0.151
10 -1.683 0.106 -1.674 0.108
11 0.282 0.781 -1.526 0.141
12 -0.121 0.905 -1.385 0.179
13 -0.685 0.5 -1.509 0.145
14 -0.554 0.585 -1.548 0.135
15 -1.164 0.256 -1.67 0.108
16 2.101 0.047** -1.42 0.169
17 -0.157 0.877 -1.392 0.177
18 -1.478 0.153 -1.527 0.14
19 -0.463 0.647 -1.574 0.129
20 -1.653 0.112 -1.712 0.1*
***, **, * represents significance at the 0.01, 0.05 and 0.1 level respectively
The T-test results in Table 4 have much more consistent evidences to the hypothesis 2 than hypothesis 1 in Table 3. Similarly, the market has reaction to the decreased dividend news earlier than the announcement date. On day -8, the abnormal return stated is the most significant that T possibility is lower than 0.01 with -3.417 T-value; day -7 described a less significance trend in level 0.05 with -0.974 maintaining negative trend until day -2 which is significant at level 0.1. After the negative news published to the market, the T values remain below 0 and significant on day 7. The interesting thing is that on day 16, the abnormal return become positive at the 0.05 significance level which shows the abnormal reaction of market.
To the cumulative abnormal return on right hand side, the decreased news has significant effect in four days ahead. Specifically, the CAR of day -4 and day -3 are negative in regard of T value in 10% significance level (-1.799 and -1.918). Furthermore, a substantial t possibility of 0.021 in day -2 indicates a more negative T-value (-2.488) of CAR which is closer to the event date. Until the end of event window, there is no positive t-value any more, and the CAR of day 20 is significant with -1.712 t-values.
In summary, the T-test results of decreased dividend news are more indicative to the hypothesis 2 than hypothesis 1. The reasons behind them may be explained by the market interpretations. Firstly, positive dividend news is not beneficial to all the investors. The increased dividends may be financed by the interests of minor shareholders if the performance activities of company did not match the expectation of managers. Secondly, the evidences showed have proved that stock market is more sensitive to negative news under the economic regression. Bad news contains more valuable information than good news which would modify the preference of less powerful investors. Lastly, the cumulative abnormal returns discussed here base daily return, and different periods should be considered at the same time for comprehensive understanding. Therefore, the following test would make assumption to provide evidence to all kinds of news for hypothesis 3.
3.2.2.3 Hypothesis 3: The market value (stock price) of Company would fluctuate significantly when new dividend announcements are made.
According the method description in formal section, table 5 below is listed for analysing the daily abnormal returns and selected cumulative abnormal returns in event period no matter good news or bad news. The detail statistics can be found in Appendix 6. The original statistics are all obtained from the estimation of market model of 39 sample companies and mean daily return is based on the aggregation of certain companies and trading days.
Table 5 describes the daily abnormal returns from trading day -10 to 10 in column A and 11 selected cumulative abnormal return periods in column B. The selected event windows are the sub cumulative abnormal returns during the 40-day event period. These different periods are considered as having high possibility that would contain more abnormal returns before or after the dividend announcement date.
Table 5: The Stock Market Reaction to Dividend Announcements (See Appendix 6)
Day A. Daily abnormal returns (AR) B. Cumulative Abnormal Returns (CAR)
AR T sig.(2-tailed) Period CAR T sig.(2-tailed)
-10 -0.00241 -0.69 0.496 (-20,-15) -0.0052336 -1.076 0.289
-9 0.00271 0.817 0.419
-8 -0.00516 -2.54 0.015** (-20,-10) -0.0018517 -0.256 0.799
-7 -0.00364 -1.21 0.235
-6 -0.00062 -0.23 0.816 (-10,-1) -0.0163486 -2.407 0.021**
-5 -0.00650 -2.68 0.011**
-4 -0.00149 -0.6 0.553 (-1,+1) 0.01090708 1.17 0.249
-3 0.00160 0.529 0.6
-2 -0.00469 -1.94 0.06* (-1,0) 0.01165495 1.353 0.181
-1 0.00385 0.997 0.325
0 0.00780 0.918 0.364 (-2,+2) 0.00269878 0.27 0.788
1 -0.00075 0.324 0.748
2 -0.00352 -1.36 0.181 (-5,+5) -0.0045081 -0.355 0.725
3 -0.00183 0.619 0.54
4 0.00126 0.375 0.71 (+5,+15) -0.0172018 -1.746 0.089*
5 -0.00024 -0.08 0.934
6 -0.00140 -0.55 0.588 (+10,+15) -0.0067194 -0.889 0.38
7 -0.00542 -1.67 0.102
8 -0.00643 -2.13 0.04** (+10,+20) -0.0023797 -0.192 0.848
9 0.00301 1.039 0.305
10 -0.00293 -1.02 0.315 (+15,+20) -0.0006789 -0.092 0.927
***, **, * represents significance at the 0.01, 0.05 and 0.1 level respectively
Among these 21-day daily abnormal return, it is observable that only 6 out of 21 trading days have positive returns and especially days -1 and 0 have significant deviation from others. Turning to the null hypothesis, there are 4 trading days which have significant differences. On day -8 and -5, the abnormal returns are higher than expected and different from 0 at level 0.05. However, the daily returns around announcement date do not as abnormal as expected and only have 0.06% with a T value of -1.94 on day -2 which is significant in 90% confidence interval. After the announcement day, day 8 have negative wealth effect in level 0.05. So from the one sample T test result of column A, it could be concluded that there is no significant difference between the sample mean and 0 except day -5, -8, -2 and 8. The market reaction of dividend policy is stable and slight on London Stock Exchange Market.
In the view of column B for analysing cumulative abnormal return, there is no heavily significant difference between mean CAR and 0. To the CAR itself specifically, window (-1, +1), (-1, 0) and (-2, +2) have positive abnormal return and the CAR on day (-1, 0) placed firstly as which having the highest unexpected return. Besides of these three windows, all the other windows are stating negative returns. Moreover, under the T test hypothesis, window (-10, -1) has the most significant wealth effect in 0.05 level during the whole event windows, which shows 0.021% possibility with -2.407 t value. Before this period, there is no information leakage during day -20 to -10; and the market reaction of dividend news is not significant until event (+5, +15) that have 0.089% expectation in t=-1.746. Therefore, abnormal returns are not significant especially 11 days before or 15 days after the dividend announcement date.
From the parametric test, the statistics show that the market have reaction to the announcement of dividend but not as significant as expected. The daily returns have been affected around 8 days before event day and the event windows also prove that cumulative returns before the announcement date is even higher than windows after announcement. However, the results in Table 5 are supportive to the hypothesis 3 that dividend news did have signalling effects to the stock price fluctuation.
Comparing with the previous empirical study about dividend policy, the outcomes of parametric test research are not as obvious as other researchers. The limitations of this research may on account of test samples. Due to some omission about the separation of cash dividends and stock dividends in public information, the results showed are kind of mixed. Theoretically, cash dividend news may have been affected by tax factors, the mixture of these two forms of dividend could lead to the insignificance of the test. Furthermore, the industry differences and sample size have to be considered at the same time. In order to improve the usefulness of this research, it is necessary to discuss the AR by non-parametric test as well and comparing five industries for better understanding the results.
3.2.3 Measuring and Analysing by Non-Parametric Test
Because the main objective of this research is focusing on the dividend policy effects to the abnormal returns, it could assume that there is a significant difference before-after the announcement day by non-parametric test. The sample groups here can be separated by announcement day which means that group 1 is day -20 to day -1 as before event while group 2 is day 0 to day 20 as after event. The result can be attained by Mann-Whitney U test which is a most well-known non-parametric significance tests. If the significance levels have reached 0.1, it could conclude that one of two samples tends to have higher values than the other one. Otherwise, the group median is considered no difference and the dividend announcements do not affect the market value.
Table 6: (1) U-test result to the positive increased dividend group
(2) U-test result to the negative decreased dividend group
(3) U-test result to the combined of three dividend groups
Source: Compiled by author from SPSS 2012/8/8
The result showed in above tables describes the basic group samples: there are 20 data in group 1 and the mean rank are 18.45, 19.7 and 20.3, while there are 21 data in group 2 whose mean rank are 23.43, 22.24 and 21.67 respectively. The test statistics give evidence that the two sample groups have higher possibility than 0.1 significant levels (significance in 0.183, 0.498 and 0.715) in same median. That is to say, there is no significant difference between the median number of real return and estimated return in two periods, which means that H0: Median1 = Median2 is accepted while H1 Median1 Median2 is rejected. So the dividend irrelevance theory has been proved by this U test. To conclusion, the U tests for the three hypotheses imply that no matter what kinds of dividend news, the stock market is sufficiently efficient that abnormal returns are observable but would be unlikely to capture significant interests by interpreting dividend news.
3.3 Industry Analyses
After parametric tests and non-parametric tests, it is difficult to give a totally definite conclusion to the hypotheses as the results are not supportive to each other. The major reason behind it could be explained by the choice of different industries. The characteristics of different product markets may determine the combination of test results. The 39 sample companies chosen contain Food & Drug Retailers (11), Real Estate (4), Banks (5), and Gas, Water & Multiutilities (13) and Automobiles & Parts (6). All the company information is obtained from the website o
