Market Reactions to Dividends Announcements and Payouts. Empirical Evidence from the Warsaw Stock Exchange

The main goal of this paper is the empirical examination of the Polish stock market reactions to dividend announcements and dividend payouts made by the companies listed on the Warsaw Stock Exchange (WSE). The research sample comprises 56 companies (WIG index constituents) that announced dividend payments and completed the payout during 2013. In the analysis, event study methodology is employed including either calculating abnormal returns and cumulative abnormal returns around the event day or testing their statistical significance using parametric and nonparametric tests. The average cross-sectional abnormal return calculated for the entire sample is found to be significant on the dividend announcement day (t = 0, 0.86%) and on one day after (t = 1, 0.59%) at the 1% and 10% significance levels, respectively. The outcomes of the analysis conducted within the three distinguished subsamples are rather more diverse. In the subgroup of the first announced dividends (or dividends announced after a minimum one-year break), the significant average abnormal return is found on day t = 1 (0.90%, 5% significance level), whereas in the case of the dividend decreases subsample, the significant average abnormal returns (at the 10% significance level) occur on days t = −4 (-1.44%) and t = 2 (-1.15%). The average abnormal return calculated within the subsample of dividend increases turns out to be positive and significant on day t = 1 (1.03%, 10% significance level). The results obtained for the average cumulative abnormal returns corroborate the findings reached for the average cross-sectional abnormal returns in the case of the first dividend and dividend increase subsamples. However, the average cross-sectional abnormal returns calculated within the eleven-day-long event window around the dividend payment day turn out to be statistically insignificant. The obtained results provide evidence that the Polish stock market reaction to dividend announcements is positive and immediate. However, the market does not significantly react to dividend payouts, which may lead to the conclusion that the WSE directly incorporates news on dividends into stock prices. Moreover, the reaction of the market for dividend announcements is consistent with the sign of the dividend change: dividend-increase (-decrease) announcements are interpreted as a positive (negative) signal by the investors. Such results support both the informational content of the dividend hypothesis and the dividend signaling hypothesis. Considering that the observed abnormal market behavior disappears within two days at most after the announcement date, the results of the study can be useful for financial practitioners only with regard to short-term investment decisions.

0.59%) at the 1% and 10% significance levels, respectively. The outcomes of the analysis conducted within the three distinguished subsamples are rather more diverse. In the subgroup of the first announced dividends (or dividends announced after a minimum one-year break), the significant average abnormal return is found on day 1 + = t (0.90%, 5% significance level), whereas in the case of the dividend decreases subsample, the significant average abnormal returns (at the 10% significance level) occur on days 4 − = t (-1.44%) and 2 + = t (-1.15%). The average abnormal return calculated within the subsample of dividend increases turns out to be positive and significant on day (1.03%, 10% significance level). The results obtained for the average cumulative abnormal returns corroborate the findings reached for the average cross-sectional abnormal returns in the case of the first dividend and dividend increase subsamples. However, the average cross-sectional abnormal returns calculated within the eleven-day-long event window around the dividend payment day turn out to be statistically insignificant. The obtained results provide evidence that the Polish stock market reaction to dividend announcements is positive and immediate. However, the market does not significantly react to dividend payouts, which may lead to the conclusion that the WSE directly incorporates news on dividends into stock prices. Moreover, the reaction of the market for dividend announcements is consistent with the sign of the dividend change: dividend-increase (-decrease) announcements are interpreted as a positive (negative) signal by the investors. Such results support both the informational content of the dividend hypothesis and the dividend signaling hypothesis. Considering that the observed abnormal market behavior disappears within two days at most after the announcement date, the results of the study can be useful for financial practitioners only with regard to short-term investment decisions.

Introduction
The issue of payout policy is of great importance for companies' managers and remains one of the most interesting problems in theoretical and empirical finance. Since Lintner's (1956) and Miller and Modigliani's (1961) original papers, a number of theories, which often have divergent views on the determinants and consequences of dividend payouts, were developed and empirically investigated. A consensus, however, has not been reached, and we still do not know exactly why companies choose to pay dividends. In prior studies, the problem of stock price responses to dividend announcements has attracted particular attention. Regardless of the predictions of the Miller and Modigliani (1961), dividend irrelevance hypothesis, the previous studies on the developed markets confirm that stock markets do react to dividend announcements. Thus, dividends convey valuable information for shareholders about the future prospects of the companies and can be a valuable tool used by managers to signal the financial condition of the companies.
Although these questions of fact have been the subject of the extensive research on mature markets, less attention has been paid to the emerging stock markets. The literature regarding those markets is rather limited, and the conclusions remain ambiguous. Thus, it is justified to investigate dividend announcement and payout effects on the behavior of the Polish stock market.
The aim of the paper is to investigate the stock market reaction to dividend announcements and dividend payouts made by companies listed on the Polish stock market, namely, the Warsaw Stock Exchange (WSE).
The research is based on a sample of 56 companies that announced dividend payments and completed the payouts in 2013. In the analysis, the event study methodology is employed.
In addition to the main goal of the paper, the reactions of the stock market among different groups concerning dividend announcements are investigated.
Similar to other studies, groups were distinguished by comparing the dividend level in 2013 to the previous year. Thus, the sample was divided into three groups: a) first dividend announcement or announcement of dividend resumption after a minimum one-year break, b) dividend-decrease announcement, c) dividend-increase announcement.
In the present paper, we make three contributions to the existing literature on the Polish stock market.
First, we treat the first official recommendation containing information of the dividend amount as the date of announcement. The present study differs from prior studies that equated the dividend announcement day with the date of the annual shareholders meeting. Second, we concentrate only on the effect of dividend announcements by clearing the dataset and removing the earning announcements. Third, we not only employ common parametric tests to verify the results, but we also recognize the distribution of the stock returns and introduce nonparametric tests when needed.
The remainder of the paper is organized as follows.
In section 2, a concise literature review regarding the impact of dividend announcements and payouts on stock prices is presented. In section 3, the Polish stock market is briefly characterized, and results from previous studies conducted on companies listed on WSE are presented. In section 4, the outline of empirical design is described, while in section 5, the results of the research are discussed. In the last section, the conclusions and implications for further research are presented.

Literature review
The relevance or irrelevance of dividend policy to the market value of shares has been one of the most discussed topics in the financial literature since the pioneering works of Lintner (1956) and Miller and Modigliani (1961). In the latter paper, the reasoning is conducted under three important assumptions: perfect capital markets, investors' rational behavior and investors' perfect certainty. The authors conduct the analysis for a single type of financial instrument, namely, stocks (Miller & Modigliani, 1961), and prove that with a given investment policy, the dividend policy cannot influence either the company's valuation or the shareholders' total return (the dividend irrelevance hypothesis). In Miller  dividend policy to the date of the announced dividend change, the investors might perceive this shift as a consequence of changing management views of company's future earnings and growth opportunities. Thus, the shift in the dividend ratio provides an opportunity to change the price of shares. Still, the market valuation is entirely based on company's investment policy and growth opportunities.
Contrary to Modigliani (1961), Lintner (1956;1962) and Gordon (1959) argue that investors differentiate between capital gains and dividends ('bird in hand fallacy hypothesis'). Dividends are not irrelevant for a firm's value since in investors' perception, dividends tend to be less risky in comparison to 'uncertain' capital gains. Moreover, interviews with managers of 28 companies brought Lintner to the conclusion that dividend policy plays an essential role in managers' decisions. In their responses, managers formed strong beliefs about investors preference over a stable rate of dividend payouts and the positive market reaction for stable or gradual growth in the dividends rate. Managers were also more reluctant to cut than to raise dividends in response to earnings changes and avoided making changes in dividends ratio that were to be soon reversed.
Lintner's findings inspired the discussion among researchers, and his partial-adjustment model of dividends was also under investigation in signaling theories of dividends, which recognize the capital market's imperfections, particularly information asymmetries.
As Bhattacharya (1979) indicates, outside investors have imperfect information about a company's profitability and future cash flow. Hence, managers having this 'insider' knowledge can use the dividend as a signal and inform the market about their expectations regarding the financial condition of the company (Dasilas & Leventis, 2011). The dividend-signaling hypothesis implies that an increase (decrease) in dividends positively (negatively) influences stock prices (Baker, Powell, & Veit, 2002). Developments of dividend signaling models can be found in works of Bhattacharya (1979), John andWilliams (1985), Miller and Rock (1985), Myers and Majluf (1984), Ofer and Thakor (1987).
Positive investor reactions in response to a dividend increase are also expected in agency theory. Here, dividend payments solve the problem of potential overinvestment since dividends reduce the level of available free cash flow that could be used by managers to invest in less profitable projects. In this view, managers having higher levels of free cash flow at their disposal induce the firm's growth beyond an optimal size (Jensen, 1986). A major contribution in the development agency theory comes from the works of Easterbrook (1984), Jensen and Meckling (1976), Jensen (1986).
Based on the company's life cycle theory, Grullon, Michaely, and Swaminathan (2002) confirm the occurrence of the negative relationship between the investment level and dividends and eventually conclude that more mature companies with fewer investment opportunities tend to pay higher dividends. Agency theory implies that by paying dividends, the companies are exposed to an increased level of market discipline and monitoring (Baker et al., 2002).
The fact that investors are not indifferent between receiving dividends or capital gains and reveal strong preference toward dividends is also recognized by the theories of behavioral finance. This theoretical stream goes back to the works of Shefrin and Statman (1984) and is based on the main features of the prospect theory developed by Kahnemann and Tversky (1979), also broadened in their later study (Tversky & Kahnemann, 1992). First prospect theory explains the widely-recognized phenomenon why dividend cuts have larger negative impact on stock prices that the positive impact of raise in dividends. In terms of the prospect theory, decreasing dividends are perceived as making losses (decreasing investors welfare), as such losses are having more pronounced effect than gains. The second behavioral approach to dividends recalls that an investor's assessment of the investment opportunity is relative to the chosen benchmark, and the investor undertakes such cognitive operations as 'coding' (Forbes, 2009, pp. 340-341). Among the theoretical works based on behavioral finance, the catering theory of dividends proposed by Baker and Wurgler (2004) Gajdka (2013) and Kowerski (2011).
In their more recent paper, Baker and Wurgler (2012) also rely on prospect theory findings and argue that investors assess current dividends against past dividends, which serve as a 'psychological reference point' . Furthermore, Baker and Wurgler (2012) develop a multi-period dividend model that is consistent with the empirical findings of dividend announcements (the signaling effect of dividends) and explains the observed managers' behavior of 'dividend smoothing' (Linter partial adjustment model).
The market reactions to dividend announcements were examined in numerous empirical studies, which were mostly focused on large, developed stock markets (e.g., USA). In recent years, less developed stock markets have also received greater attention, and thus, the results have become more diverse.
Most empirical studies confirm the existence of the positive relationship between dividend announcements and stock price movements. The classic Pettit's (1972) study based on a sample of 625 companies listed on the NYSE in the period of 1964-1968 proves that dividends carry valuable information. Pettit divides the dividends dataset into seven categories: omissions, reductions, no change, initial payment, less than 10% increase, 10-25% increase, 25% or higher increase, whereas the initial payment group comprised companies that paid no dividend in the previous period.
Pettit confirms that an increase (decrease) in dividends induces positive (negative) abnormal returns. It is pertinent to note that the major conclusions are drawn based on the monthly price data. Pettit additionally examines a much smaller sample of the daily stock returns (approximately equal to 10% of the monthly data sample) and finds that the market discounted the information within one day after the announcement.
The author is aware that in such studies, the effect of related market information (i.e., earning announcements) must be excluded or at least considered. However, his proposal for overcoming this problem seems to be somewhat ambiguous and does not prevent the influence of the earnings announcements on the ob- The effect of the first (initial) dividend announcement and an announcement after a long 'non-dividend' period is investigated by Asquith and Mullins (1983) on a sample of 168 companies listed on the NYSE or ASE for the period 1963-1980. The authors assume that dividend announcements are unexpected by market participants (see also Gurgul, Mestel, & Schleicher, 2003). As Asquith and Mullins report, for almost 70% of companies, there was a positive market reaction to the initial dividend announcement. Asquith and Mullins additionally control for the effects of other important events that might have influenced the results and conclude that dividends convey unique information to investors. The authors also find support for the hypothesis that dividends and earnings announcements can be partial substitutes. However, the results may be disturbed by an unequal division of the announcements into the analyzed subsamples.
The joint impact of dividend announcements on stock and bond prices is investigated by Dhillon and Johnson (1994). The authors aim at distinguishing between two hypotheses, namely: the information content (1) and wealth redistribution (2 Lonie et al. (1996) indicate that in the first six months of each year, almost all dividends are already announced to market participants. Second, the authors allude to the studies conducted by Chowdhury and Miles (1989) and DeAngelo, DeAngelo, and Skinner (1992) and highlight that mixing events of dividend announcement from different periods might be inappropriate, since the interpretation of dividend signals is influenced by economic conditions. Generally, the results favor the statistically significant positive reaction of share prices in a 2-day window in the case of a dividend increase and negative abnormal returns in the case of dividend cuts (Lonie et al., 1996). Moreover, Lonie et al. (1996)  There are also several studies devoted to the smaller European stock markets located in Austria, Ireland and Greece. Gurgul et al. (2003) investigate the Austrian stock market reaction to dividend change in terms of prices and trading volume. In this paper, there are 181 first dividend announcements recognized, which are defined -contrary to the previous studies -as 'the very first official statement on dividends of the executive board' (Gurgul et al., 2003 (Gurgul, 2012).
The same definition of dividend announcement as Gurgul et al. 's (2003) is also adopted in the Dasilas and Leventis (2011)

Evidence from the Warsaw Stock Exchange
The history of the Polish stock market is rather short since it was only re-established in 1991 together with market reforms introduced in Poland (Ziarko- Siwek, 2008, p. 347). The main motive prevailing in the construction of the WSE was to enable smooth privatization processes to aid in effective capital allocation, as Janicka (2005) notes. Since its establishment, the WSE has experienced growth in terms of capitalization and number of companies listed, and currently is a recognized regional stock market (see also Mrzygłód & Nowak, 2013). However, it is worth noting that the WSE remains a relatively small market  Gurgul and Majdosz (2005) confirm the positive reaction of the stock market (+0.79%) on the day after news release (t+1). Contrary to studies on the US and European markets, Gurgul and Majdosz (2005) assume that any announcement concerning dividend payment should be treated as positive information for investors. Additionally, the authors examine the stock price reaction of the company's rivals coming from the same sector of the economy and reveal that industry rivals also experience positive stock price movements on the second day after announcement (t+2).
The larger sample of dividend events (245) is employed in the Tuzimek's study (Tuzimek, 2013, pp  In the case of stable dividends, the abnormal returns are found to be positive and statistically significant on the day of the announcement (0.76%) (Tuzimek, 2013, p. 279-281). Additionally, the author investigates the effect of dividend announcements on stock prices dividing the companies into selected categories, such as company size, turnover liquidity, P/E ratio, P/BV ratio, ROE, debt ratio.

Data and methodology
The research sample contains companies listed on the WSE selected based on two criteria: (1)  In this study, the standard event study methodology is employed. The calculation is conducted based on daily data of stock prices derived from the official WSE website, gpwinfostrefa.pl. The actual daily stock return for i company is calculated as follows: The basic event period comprises 11 days around the dividend announcement (or payment) date, which gives the event window equal to (-5, +5) is related either to the problem with time-varying parameters in the market model or to the high probability of not fulfilling the classical OLS assumptions in the case of using the daily data in estimation of the model parameters. However, computing abnormal returns according to equation (2) does not require using the estimation window, which eventually excludes the possibility to improve the properties of selected parametric tests, which will be discussed in detail below.
The average cross-sectional daily abnormal return is computed as follows: where N indicates the number of observations used in the study. The null hypothesis maintains that the abnormal return on day t within the event window is equal to zero: which means that the considered event has no influence on stock prices behavior. This hypothesis can be tested using the parametric test based on the ratio of cross-sectional mean abnormal returns and the stan- Assuming that the abnormal returns are independent, identically distributed and normal, the t-statistic has a Student-t distribution with where N indicates the number of observations used in the study.
However, daily returns (either actual or abnormal) are, in general, not normally distributed; according to the standard central limit theorem, the cross-sectional mean excess return converges to normality with the increase in the number of securities. Otherwise, the usage of statistic (5) has a different limitation: the event occurrence on day t usually leads to the contemporaneous changes of the numerator and denominator of the expression (5). Thus, it can lead to the situation when the t-statistic remains statistically insignificant, although the event considerably determines the stock prices (Gurgul, 2012, p. 51). Regrettably, equating abnormal returns to market adjusted returns instead of excessive returns over the market model impedes either employing the standard deviation of abnormal returns computed for estimation window (Gurgul, 2012, p. 51) or calculating the standardized abnormal returns and using parametric test proposed by Boehmer, Masumeci and Poulsen (1991).
Taking the asymmetry of the cross-sectional abnormal return distribution into account, the nonparametric Corrado's rank test (Corrado, 1989) Under the null hypothesis expressed as (4), statistic ) (u T is distributed asymptotically as unit normal, The generalized sign test is used to check whether the abnormal returns are independent across stocks. Under the null hypothesis of no abnormal performance, the number of positive and negative abnormal returns equals to 50 percent in the event window. The number of non-negative values of abnormal returns has a binomial distribution with parameter p (Brown & Warner, 1980;Cowan, 1992). The statistic for the sign test is defined as follows: where 0 p denotes the observed fraction of positive returns in the event period. Under the null hypothesis,

Results
The results describing abnormal returns performance within the event window (-5, +5) days around the dividend announcement day in the entire sample are summarized in Table 1 Shapiro-Wilk statistics reveal that the daily stock returns on days within the event window are not normally distributed. Table 2 contains the summary of abnormal returns behavior within the event window around the dividend payment day.
In this case, mean cross abnormal returns are found to be statistically insignificant for all 10 days around the day the dividend was paid. The results of the Shapiro-Wilk test represent significant departures from normal distribution of majority abnormal returns within the event window. The cumulative abnormal return (CAR), which is calculated in the event window around the dividend announcement day, is approximately 3.5 times as large as the CAR calculated around the dividend payment day.
The results obtained for the average abnormal returns calculated within the entire sample are con-        The results of the average cumulative abnormal returns' analysis confirm -in most cases -the findings obtained for the average cross-sectional abnormal returns. The reaction of the market is statistically significant in the daily intervals (0,1) and (0,2), either in the case of the first dividend (dividend resumes after a break) announcement (1% significance level) or in the case of dividend-increase announcements (10% significance level). In the case of dividend-decrease announcements the average cumulative abnormal returns remain statistically insignificant. Therefore, the results obtained for cumulative abnormal returns in the subgroup of dividend decreases stay in contrast to the findings reached for the average abnormal returns (compare Table 5).
The presented results generally stay in line with the other studies conducted on the Polish stock market.
In comparison to Gurgul and Majdosz (2005) The obtained results are generally in line with US and European studies (Dasilas & Leventis, 2011; i.e., Gurgul et al., 2003;Lonie et al., 1996), which proves the existence of the dividend announcement signal-  ) around the dividend announcement day in three subsamples Note: A -first dividend/dividend resumes after a break, B -dividend decreases, C -dividend increases, *, ** -significance at the 10%, 5% significance level.
Market reactions to dividends announcements and payouts. Empirical evidence from the Warsaw Stock Exchange This work is licensed under a Creative Commons Attribution 4.0 International License.
ing effect not only in the case of the news release that promises investors an increasing dividend.

Summary and conclusions
The results obtained based on 56 dividend announcements among companies listed on the WSE in the year 2013 are mixed. Based upon the sample of 56 firms, the market reaction turns out to be statistically significant and positive only on the dividend announcement day and one day after the announcement. The mean cross abnormal returns are found to be insignificantly different from zero for all ten days around the dividend payment day. Thus, the effect of dividend announcement is reflected in stock prices immediately.
Considering the direction of changes in the expected dividend payouts, in the case of increasing dividends and dividends paid for the first time (or dividend resumes after a break), the impact of the news release on stock prices is statistically significant, positive and noticed on the first day after the announcements. Furthermore, in the case of dividend decrease announcements, their impact on stock prices is negative and noticeable on the second day after the announcement.
The outcomes of the study allow us to confirm that the effect of dividend announcements is in line with the informational content of the dividend hypothesis as well as with dividend signaling models. The reaction of the market is consistent with the direction of the dividend change: dividend-increase (-decrease) announcements are interpreted as a positive (negative) signal by the investors. Moreover, the stock market reaction on the news release turns out to be rather quick.
Thus, the prices seem to 'digest' the information immediately. Based on the employed sample the authors cannot confirm that behavioral models of dividends have a good explanatory power in the case of the Polish stock market. The reaction of the stock prices, although correct in sign, has not been stronger in the case of dividend cuts compared to dividend increases.
Finally, the authors would like to add two general remarks. First, the Polish stock market is not semi-strong informationally efficient in a given period. However, it is pertinent to note that the observed abnormal market behavior disappears within two days at most after the announcement date. Second, the obtained results can be useful for financial practitioners. The statistical significance of abnormal returns in response to dividend announcements can be used to build portfolios with dividend companies. However, the profitability of such portfolios should be checked in a long-term period, which is beyond the scope of this study.