Analysis of the dividend announcement effect in different market phases

(1)

Bachelor Thesis Economics and Business Economics Faculty of Economics and Business

Academic year: 2019-2020

Analysis of the dividend announcement effect in different market phases

Comparison of the dividend increase announcements in bull and bear market phases in France

Specialization: Finance

Name: Catherine Pamela Torres Cunalata Student Number: 11644753

Supervisor: Drs. P.V.Trietsch, M.Phil. Date: 30-06-2020

(2)

Statement of Originality

This document is written by Student Catherine Pamela Torres Cunalata who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Abstract

This research analyzes the validity of the signaling theory and the differential in stock price reaction to dividend increase announcements between market phases. That is, it will examine if the price reaction to dividend increase announcements is higher in bear phases compared to bull phases. The reason is that bull phases are characterized by rising prices and low volatility, while bear phases are the opposite. Thus, this study performed an event study using the Single Index Market Model to calculate abnormal returns. It was found that the market phases significantly affect the size of the reaction around the dividend announcement days. Specifically, the size of the difference between bull and bear market phases is statistically significant on the event day. Also, this study found significant evidence for the signaling theory, which advocates that dividends carry some information relevant for firm valuation. Therefore, the results indicate that investors earn abnormal returns around the announcement day.

(3)

Acknowledgments

The author would like to thank the institution named Secretaría de Educación Superior, Ciencia, Tecnología e Innovación (SENESCYT) for the financial support provided to fulfill the bachelor program of Economics and Business Economics.

(4)

1. Introduction

Several papers such as Michaely et al. (1995), Kato and Lowestein (1995), and HuZuguang and Ahmed (2010), found evidence that an announcement of a dividend increase or decrease has an impact on the stock prices. The theories that help to explain the existence of this effect are the Signaling Theory, Catering Theory, and Free Cash Flow Hypothesis. According to these theories, dividends can convey information regarding the expected profitability of the firm or also aid in mitigating agency costs. The magnitude of the effect is measured by the abnormal returns generated around the event window.

Authors such as Michaely et al. (1995), and Jin (2000) identified that firms that decide to increase their dividend will experience positive abnormal returns, while firms that choose to decrease dividends will show negative abnormal returns. However, Below and Johnson (1996), Akron (2011), Akintade and Olade (2014) found that the size of this effect is different for different states in the economy, such as bull and bear market phases. Bull phases are periods with generally rising prices and low volatility contrary to bear phases. Therefore, they suggest that the size of the effect in bull market phases is smaller compared to bear market phases. This is because investors were already expecting a dividend increase due to their inflated expectations during bull phases as opposed to bear phases.

The motivation of this research is that often papers concentrate on determining dividend announcement effects in either recession or not recession periods (Akintade & Olade, 2014). However, there is little research done that is focused on the market phases, which might or not be a recession. Also, previous research regarding the dividend announcement effect in different market phases is mainly focused on the USA market rather than European countries. Therefore, this research tries to answer the question: Does the dividend announcement effect change in different market phases of the economy in a developed European country, namely France?. Thus, aiming to increase the external validity of previous research done in developed countries (USA,

(6)

2

UK, Israel) in order to help managers to make better dividend policy decisions and for investors planning portfolios in line with the market phase.

This research will use the FactSet database to get the monthly stock market index for France CAC 40 from 2002 to 2019 and includes both bull and bear market periods. The different periods are identified using a nonparametric methodology, which is based on rules. Thus, to determine differences in market periods, one has to check its minimum duration and if the price falls more than 20% or rises more than 15% (Edwards et al., 2003). After that, an event study will be performed in other to find and test for significance (using a t-test) the abnormal returns around the dividend increase announcement days. Therefore, to calculate the normal returns, the market model will be used with the French firms’ daily information extracted from the Compustat database and the French CAC 40 daily market index from the Factset database. Finally, a paired t-test is needed to compare the distinct reaction size between market phases.

The structure of the paper is as follows. Section 2 contains a literature review that discusses the main definitions behind the dividend announcement effect and different options for measuring variables. Section 3 provides the data and model that will be used to test the hypotheses, also the results found by performing the event study. Lastly, section 4 presents a summary of the main findings, limitations of this paper, and recommendations for future research.

2. Literature Review

This section will describe the concepts and measurements behind a dividend announcement effect. Also, it will detail The Signaling Theory, Catering Theory, and Free Cash Flow Hypothesis, which potentially explain the existence of a dividend announcement effect. After that, the definitions for bear and bull market phases will be detailed as well as the different ways to measure market phases. Moreover, some evidence regarding the influence that the market phase has on the size of the effect will be presented.

(7)

3

The following conceptual framework provides an overview of the theories that explain the effect of a dividend announcement increase in stock prices. Also, it shows that market phases influence the size of the effect produced.

Figure 1. Conceptual Framework

2.1. Definition of the Dividend Announcement Effect

The dividend announcements effect is the positive or negative effect on stock prices that occurs when management decides to increase or decrease dividends (Abdullah, Rashid & Ibrahim, 2002). Some papers support the existence of a dividend announcement while other authors do not.

Miller and Modigliani (1961) said that in perfect capital markets, the choice of dividend policy is irrelevant since it does not affect the stock prices. They assert that the benefit of an increment in dividend payments will cancel out due to the resulting capital loss for existing shareholders. Thus, they imply a tradeoff between income received now and the lower selling price.

However, in reality, markets are not perfect. Thus, several authors such as Van Horne and McDonald (1971), Kato and Lowestein (1995), and HuZuguang and Ahmed (2010), incorporated

(8)

4

market frictions in their studies. They found a significant relationship between the change in dividend payments and the change in stock prices. They were thus suggesting that the choice of dividend policy is relevant for firm valuation.

Despite the findings against the existence of a dividend announcement effect, the majority of the studies support that this effect occurs. This is because markets are not perfect, so incorporating these frictions in the analysis provides significant results. Thus it can be implied that an announcement of a dividend increase or decrease affects stock prices.

2.2. Measurement of the dividend announcement effect

In order to measure the effect of an event, an event study is performed, which makes use of the abnormal return. The abnormal return is the difference between the actual return and the normal return of a firm generated over an event window. The normal return is the expected return that would be obtained if the event did not occur. This means that the expected return is the return that an investor anticipates given the potential outcomes disregarding the option of a dividend change. After that, the abnormal returns are tested using a test statistic for significance different from zero during the event (Strong, 1992).

For example, a dividend announcement increase gives a signal to the market, which is measured by the positive abnormal return that it produces. Contrary, a dividend announcement decrease is quantified by the negative abnormal returns that it generates (Suwanna, 2012). However, how the normal return is calculated varies across papers. The following section explains the different possible alternatives to compute normal returns.

(9)

5 2.2.1. Models that measure normal returns

Some approaches are available to calculate the normal return of a given security. MacKinlay (1997) describes that the approaches can be loosely grouped into two categories: statistical and economic. The statistical models rely on statistical assumptions regarding the asset return’s behavior. On the other hand, economic models are built upon assumptions about investor’s behavior. The Figure 2 below shows the previously mentioned classification; however, this analysis will outline the most used models for an event study. A study performed by Holler (2012) found that out of 400 event studies, 79.1% adopted the market model, 13.6% the market adjusted model, 3.3% the constant mean return model, 3.6% the multi-factor models and 0.7% the CAPM.

Figure 2. Expected return models classification

Economic models such as Arbitrage Pricing Theory (APT) and Capital Asset Pricing Model (CAPM) have the advantage of calculating more accurate measures for normal return since they incorporate statistical and economic assumptions. However, CAPM is very sensible to the

(10)

6

restrictions imposed by the model itself. Although APT can solve the biases produced by CAMP, findings suggest that APT behaves like the market factor, so adding more variables do not contribute to the increment in the explanatory power. Therefore, statistical models are preferred for event studies since they can also remove these biases by assuming that returns are jointly multivariate normal, independently, and identically distributed (MacKinlay, 1997).

The most common choice between statistical models is the constant mean return model and the market model. The first one assumes that there is a constant mean return for a security over time. The second one establishes a linear relationship between the security return and the market return. The market model shows an improvement compared to the constant mean return model since it eliminates the fraction of the return that relates to the market return. Thus, reducing the variance of the abnormal return and increasing the capacity to find the effect of an event. Likewise, the market model relies on the distributional OLS assumptions, which are enough for a model to be appropriately specified. Therefore, the 𝑅2_{of the model will determine the benefit obtained from} using it (MacKinlay, 1997).

Even though economic models are more precise when calculating normal returns, they can produce biased results. Therefore, the statistical models are usually chosen since their distribution assumptions eliminate these biases. Especially, the market model is preferred due to its ability to detect the effect in an event study since it controls market affects.

2.3. Theories

Multiple prevailing theories are trying to explain the dividend announcement effect; among these are The Dividend Signaling Theory, Catering Theory, and Free Cash Flow Hypothesis. This section will present the definition of the theories, as well as the papers that found evidence in favor and against them. The differences and similarities between the theories will be discussed at the end. The following table presents an overview of the papers that relate to the analyzed theories.

(11)

7

Evidence Author(s) Country Time Variables Results Conclusion

Sig na lin g T he o ry In f a v o r Pettit (1972) USA 1967-1969 -NYSE index -dividend increases and earnings performance 10% significant dividend effect on the announcement month Market participants use the information implicit in dividend announcements Bernheim and Wantz (1995) USA 1962-1988 -NYSE/ AMEX and NASDAQ data -taxes and dividends 1% significant relationship between dividend tax and price response

The size of the abnormal returns are more sensible when changes in dividend taxation occur

Ag

a

ins

t Watts (1973) USA 1946-₁₉₆₇ -earnings and _dividends No significant _{results that prove a}

relation between dividends and stock returns

Any information conveyed by dividends is covered by the noise in the model Ca ter ing T heo ry In f a v o r Baker and Wurgler (2004) USA 1963-2000 -NASDAQ data -dividend payments and omissions Dividend initiation rate is significantly positively related to dividend premium Companies pay dividends when investors put higher prices in dividend-payer firms Mamunur et al. (2013) Malaysia 2002-2007 -non-financial firms listed on Bursa Malaysia -dividends and stock returns Market demands significantly (1%) influence dividend payments

Dividends per share influence firm value due to investor sentiment Ag a ins t Neves et al. (2006) 12 Eurozone countries 1986-2003 -growth of capital goods prices, long and short term interest rates on debt

1% significant effect after the inclusion of the interaction variable of highly liquid assets times catering

The catering effect found only in firms with highly liquid assets and good investment opportunities F re e Ca sh F lo w H y po thes is In f a v o r Jensen (1986) Review of several papers done by the author 1986 -debt, dividends, and takeovers Leverage and dividends increase stock prices around 2.2% after two days

Debt, dividends, and takeovers reduce agency costs produced by excess cash

Park and Jang (2013) USA 1995-2008 -Compustat database for the restaurant industry -debt, leverage, free cash flow 1% significant results that support better company performance by reducing the free cash flow available

Free cash flow has a direct relation with firm performance deterioration, implying an over-investment problem Ag a ins t Denis et al. (1994) USA 1962-1988 -NYSE/ AMEX database -dividends, daily returns, capital expenditures

Weak results for overinvestment tendency, but 1% significant results for cash flow signaling effect.

Find support of the cash flow signaling and stock price reactions to dividend change, but no evidence for the overinvestment hypothesis Table 1.Overview of the theories

(12)

8 2.3.1. Dividend signaling theory

The dividend signaling theory asserts that a change in dividend reflects the manager’s expectations regarding the firm’s future profitability. Therefore, a dividend increase occurs when the manager is convinced that the firm will be able to pay higher dividends in the future. In contrast, managers decrease dividends to signal that firm’s earnings might not improve (Berk & DeMarzo, 2017). Thus, investors react accordingly to these signals by generating a positive or negative stock price reaction (Lang & Litzenberger, 1989).

Pettit (1972) was the first to find evidence that supports the existence of this theory. He concluded that investors use dividends to assess the value of a security. Thus managers should be reluctant to decrease or omit dividends. Likewise, various authors such as Bernheim and Wantz (1995) and Frankfurter et al. (2003) showed the presence of positive (negative) abnormal returns after the announcement of an increase (decrease) in dividend payments. Thus, they suggest that markets are not perfect due to the existence of asymmetric information.

On the other hand, studies performed by Watts (1973) and Gonedes (1978) did not find a significant relationship between dividends and future earnings. Instead of examining the existence of an announcement effect, they focused on studying the predictive content of dividends for future firm performance. Thus they concluded that the noise of the model suppresses the information content of dividends.

After all, it can be seen that most of the evidence supports a positive relation between dividend announcement and the stock price. This is due to the existence of asymmetric information in the market. Therefore, this suggests that dividend payments or omissions carry some information and can give credible signals to investors regarding the firm’s prospects.

(13)

9 2.3.2. Catering Theory

The Catering theory states that investors have uninformed and time-varying preferences for dividend-paying stocks. Also, it says that arbitrage does not prevent that demand influences stock price between dividend payers and non-payers. Thus, managers cater to investors’ demand by increasing (decreasing) their dividend payments when investors set a high (low) price on (non-) payers shares. The catering theory argues that the propensity of firms to pay dividends can be measured by a dividend premium in stock prices, therefore indicating that dividend policy is relevant (Baker & Wurgler, 2004).

After the first paper done by Baker and Wurgler (2004), other researchers such as Brown and Cliff (2005), and Mamunur et al. (2013), found significant evidence that supports this theory. They suggest that firms pay dividends when the dividend premium for a stock is positive due to the demand. However, they advocate a disequilibrium in the market since companies pay dividends just according to the demand and not because they have income reserved for it.

Differently, Neves et al. (2006) found that the catering theory holds just for firms with good investment prospects and highly liquid assets. Besides, a study done at a global level by Ferris et al. (2009) identified that the propensity to pay dividends differs between common law countries and civil law countries.

Although some evidence does not entirely support the catering theory due to the differences in firm characteristics or jurisdictions. Several authors found a significant relation between dividends and demand, which imply that this theory holds. Thus, this suggests instability in the market since managers cater to investors’ demand for dividends, even when the possibilities for paying them are limited.

(14)

10 2.3.3. FCF hypothesis

The free cash flow hypothesis mixes features of the signaling theory and agency theory. Managers tend to invest inefficiently the firms’ funds without aligning shareholders’ interests. Therefore, debt and dividend payments help to decrease the free cash flow available to managers’ perks. Thus, this theory indicates that markets are not perfect due to the presence of agency conflicts and asymmetric information (Frankfurter et al., 2003).

Jensen (1986) was among the first to find significant evidence that supports the existence of this hypothesis. He explains that the price reaction is caused by the conflict of interests that exists between shareholders and managers. Therefore, an unexpected dividend increase payment will reduce the free cash flow that could be used for negative NPV projects (overinvestment) and raise the firm value. Moreover, Park and Jang (2013) found significant evidence that a reduction of the free cash flow available to managers will reduce agency costs, thus increasing stock value. While a decrease in dividend payments will have a contrary effect.

On the other hand, Denis et al. (1994) found significant evidence consistent with the cash flow signaling and stock price reaction. However, they could not support that a decline of free cash flows available reduce overinvestment. Thus, they imply that excessive investment cannot be reduced by decreasing free cash flow.

Although some evidence indicates that overinvestment problems are not mitigated by reducing the cash available to managers. Several authors support the existence of the free cash flow hypothesis. Thus indicating that dividend payments can reduce agency costs and provide positive signals to the market.

(15)

11 2.3.4. Comparison of the theories

After analyzing the theories, all of them discard the existence of perfect markets. They outline that market imperfections are mainly agency conflicts and asymmetric information. Thus, the above-mentioned theories suggest that dividend policy is relevant for firm valuation. Lang and Litzenberger (1989) mention that the free cash flow hypothesis needs signaling and agency theory to complement each other. Most of the evidence strongly supports the presence of the signaling theory.

Asquith and Weiss (2016), mention that although the existence of catering and free cash flow theories that explain why firms pay dividends, none work as good as signaling theory. They affirm a decline in the signal that a dividend announcement produces. This is due to the development of the Internet, which allows investors to be better informed; thus, the size of the market reaction will be reduced. Despite the decrease, there are still significant market reactions. However, the problem is that the signaling theory is not able to explain the reason why dividend-payer firms have declined over time.

2.4. Bull and bear market phases

Bull market periods are characterized for raising prices and low volatility while bear market periods are considered periods of declining prices and high volatility. Moreover, bull phases are periods with strong investor interest in the stock markets. It is also known that recessions (booms) are always associated with bear (bull) market phases, but not vice versa (Gonzalez et al., 2005).

It was already discussed that there is significant evidence showing that a dividend announcement affects stock prices. However, Below and Johnson (1996) and Wann and Lobo (2009) suggested that investor reaction to a dividend change is different in bull and bear market phase. Thus, they explain that the information conveyed by dividend announcements is smaller in bull phases

(16)

12

compared to bear phases since these changes might already be expected by investors. Similarly, Akron (2011) mentions that during a bear phase, dividends provide a credible signal of firm profitability, thus producing a stronger reaction.

Hence, this research will focus on testing the differences between the size of the dividend announcement effect in different market phases. In order to do that, the phases need to be recognized first. Therefore, the different ways to identify bull and bear market phases will be outlined in the next section.

2.5. Models that measure market phases

There are two ways to measure bull and bear market periods: nonparametric (rules-based) and parametric (econometric models based). Both approaches showed to be similar, but the differences are mainly on the greater transparency of the first against the better insight into the data generating process of the second (Harding and Pagan, 2003).

The nonparametric approach for recognizing market cycles was first used by Bry and Boschan (1971) and lately applied by Harding and Pagan (2003). This method uses rules to locate peaks and troughs over time, which will be the turning points for identifying bull and bear market periods. The identification rules relate to the minimum duration of the cycles (15 months) or if the stock price falls more than 20% or rises more than 15% (Edwards et al., 2003).

The Markov regime-switching model pioneered by Hamilton (1989) is the most commonly applied among parametric models. This method uses time-varying probabilistic inferences to determine the bull and bear market periods in an index. Thus, a bull period depicts high volatility and high return, while the bear period goes in the opposite direction. Also, the fit of the model can be improved by adding more regimes or by including individual characteristics of financial markets (Kole & Dijk, 2017).

(17)

13

The differences between the approaches have been analyzed by authors such as Harding and Pagan (2003), Lunde and Timmermann (2004), and Kole and Dijk (2017). The rules-based approach is easier to understand and execute. Also, it might be more robust since it does not need severe assumptions regarding the return distributions or variations through time. However, the outcome can be affected since the inputs required are subjective. On the other hand, the model-based approach is more complicated to understand, and the results can be hugely affected if the data generating process is different. Nonetheless, it improves statistical efficiency by dealing with identification and prediction at once, as well as the ability to expand the model to cover more states.

Although the parametric methodology improves statistical efficiency and can incorporate more states of the economy. The nonparametric approach is preferred due to the ease, flexibility, and transparency that it offers. Also, since this study needs to find market tendencies and ex-post series of bull and bear markets, the nonparametric works better, as suggested by Kole and Dijk (2017).

2.6. Effect size

Previous studies such as Michaely et al. (1995) found that there is an abnormal return surrounding the dividend announcement day. They compared the portfolios that increased or decreased dividends vs. benchmark portfolios. The results are a 1% significantly positive excess return (3.4%) for initiation portfolios, while there is a negative excess return of -7% for omitting portfolios. Morever, Hu Zuguang and Ahmed (2010) obtained 10% significantly positive abnormal returns on the dividend announcement day in the Shanghai stock exchange. Kato and Loewenstein (1995) investigated the abnormal returns generated before and after a tax reform for dividends. They also found 1% positive significant abnormal returns on the announcement day. After all, the results strongly support the existence of abnormal returns after dividend initiations or increases and dividend omissions or decreases. The findings of the previously mentioned papers are outlined in the table below.

(18)

14

Author(s) Country Time Variables Results (%)

Michaely et al. (1995) USA 1964-1988 -NYSE/AMEX -security returns, dividends -Dividend initiation: Excess return: 3.4 *** -Dividend omission: Excess return: -7 *** Kato & Loewenstein (1995)

Japan 1981-1991 -Tokyo stock

exchange -stock prices, ex-dividend days

Before tax reform -AR 0= 0.54 *** After tax reform -AR 0= 0.01 *** Hu Zuguang &

Ahmed (2010).

China 2005-2009. -Shanghai stock

exchange -Stock returns, dividends Dividend increase: -AR 0= 0.59 * Dividend decrease: - AR 0= 0.80 * Table 2. *p < .10. **p < .05. ***p < .01

However, the size of the abnormal returns mentioned before varies according to the state of the economy. Below and Johnson (1996) and Wann and Lobo (2009) found that the effect of the signal of a dividend announcement varies according to the market phase. Those authors identified a significant relationship between the market phase and the abnormal returns generated around the announcement day.

Below and Johnson (1996) suggest that investor reaction to a dividend change is different in bull and bear market phases. They performed their study analyzing pre-event abnormal returns, which are the abnormal returns five days before and including the event day, and the post-event abnormal returns, which are five days posterior and including the event day. They found that abnormal returns are significantly larger in bear phases than in bull phases pre and post the dividend announcement event. This means that during bear phases, a dividend increase produces a higher positive effect in the price compared to bull phases, which give smaller positive abnormal returns.

Likewise, Akintade and Olade (2014) support the findings of Below and Johnson (1996). They found evidence that dividends that change opposite to the market phase significantly affect stock prices after the event. Their research strongly advocates in favor of the validity of the signaling theory. Even though they could not find significant pre-event results, they found significant post-event results. Also, they mention the possible presence of insider trading since they found significant evidence (10%) for CAR (-5, +5).

(19)

15

Similarly, Akron (2011) concluded that investors’ reaction to dividend announcements in bear periods is significantly stronger than bull periods on the event day. He found a 5% significant difference between bull and bear periods on the day the dividends are announced. However, he could not find significant cumulative abnormal returns in the windows before and after the announcement event.

Overall, it can be concluded that a dividend announcement produces an effect on the stock price measured by the abnormal returns generated. Nevertheless, the size of the effect differs according to the market phase. Mainly, bear phases produce higher abnormal returns compared to bull phases when a dividend increase is announced. The table below indicates the results of the authors previously mentioned.

Author(s) Country Time Variables Results (%) T-test of difference

Below & Johnson, 1996

USA 1970-1982 -bull and bear/market return: S&P 500 index -security return: CRSP NYSE/AMEX daily closing price -pre-event AR Bull: 0.356 *** Bear: 0.462 *** -post-event AR Bull: 0.366 *** Bear: 0.457 *** -pre-event AR t-value= 2.07 *** -post-event AR t-value= 1.81 ** Akintade & Ololade, 2014

UK 1970-1982 -bull and bear/market return: FTSE 100 Reuters Database -Reuters Database daily trading share price and market returns -Day 0 AR: Bull: 0.393 ** Bear:0.494 ** -CAR (-5, +1) Bull: 0.367 Bear: 0.427 -CAR (+1, +5) Bull: 0.315 * Bear: 0.421 * -Day 0: t-value= 2.17 ** Akron, 2011

Israel 2001-2007 -bull and bear periods: cyclical market returns Tel Aviv 100

-closing prices of large cap companies from the Tel Aviv 25 index

-Day 0 AR: Bull: 0.147 ** Bear: 0.918 *** -CAR (-10,-2) Bull: 0.291 Bear: 0.056 -CAR (+1,+10) Bull: 1.221 Bear: 1.623 -Day 0: t-value= N/A ** Table 3. *p < .10. **p < .05. ***p < .01

(20)

16 2.7. Hypotheses

After analyzing the previous sections, it can be concluded that there is an effect on the stock price after a dividend announcement occurs. Thus, this paper aims to provide evidence on the size of this effect in different market phases. In order to do that, an event study will be conducted, therefore testing the following hypotheses.

Hypothesis 1: The dividend increase announcement has a positive impact on stock prices.

Hypothesis 2: Firms increasing their cash dividend will experience a bigger change in price in bear

phases compared to bull phases.

3. Empirical analysis

This section will present the data gathering process needed for this study. After that, the methodology employed will be explained. Lastly, the results obtained will be analyzed and compared.

3.1 Data

The following subsections will indicate the sources used to collect the data. The first part regards the identification of the market periods. The second part concerns the database used for collecting the inputs for the model used in the study.

3.1.1. Identification of periods based on the nonparametric perspective

In order to identify market phases with the nonparametric approach, the FactSet database was used to download the stock market index (monthly data) for France CAC 40 from 2002 to 2019. This

(21)

17

method looks for periods of a generalized upward trend, which will be identified with the bull phases and periods of a generalized downward trend, which will be identified with the bear phases. It tries to locate the turning points, which are the points at which the trend changes direction. Those are recognized by the peaks (highest point in the trend) or troughs (lowest point in the trend) in the series. Therefore, the following steps are applied:

1. Location of the turning points (peaks or troughs) in the market index, which signal the change in the market trend.

2. As suggested by Pagan and Sossounov (2003) this paper uses the natural logarithm of the stock price 𝑝_𝑡 over a window width of 8 months. Thus it recognizes a peak when [𝑝𝑡−8, … , 𝑝𝑡−1< 𝑝𝑡 > 𝑝𝑡+1, … , 𝑝𝑡+8] and a trough if [𝑝𝑡−8, … , 𝑝𝑡−1< 𝑝𝑡 > 𝑝𝑡+1, … , 𝑝𝑡+8]

The following criteria are included in order to avoid the identification of spurious phases:

1. Turns that occurred within the first/last eight months of the series are eliminated.

2. Bull or bear phases that last less than 16 months are eliminated.

3. Phases of less than four months are eliminated unless the increase/decrease is more than 20% (rule of thumb for identifying market phases).

4. Implement alternation by taking the highest/lowest of two consecutive peaks/troughs, which ensures that a peak follows a trough or vice versa.

3.1.2. Data needed for the market model

The sample of companies used in this study was obtained from the Compustat database Global- Security Daily (French companies) from 2002 to 2007 and 2012 to 2016. The variables downloaded are gvkey (unique company identifier), company name, closing price daily (in

(22)

18

dollars), cash dividend daily (in dollars), announcement date (dividend declaration date), gsector (Global Industry Classification Sector). Also, the market return was calculated with the index prices (in dollars) extracted from the Factset database (daily data) for France CAC 40 from 2002 to 2007 and 2012 to 2016.

Authors such as Below and Johnson (1996), and Akintade and Ololade (2014) found that the data could have sampling bias. This bias results from including firms that initiate or resume dividends just before the start of a market phase, or firms that discontinue dividends right after the market phase finishes. Therefore, in order to avoid it, a firm was included in the analysis only if it paid continuous dividends two years before and two years after each phase. Thus, the sample will contain companies that give investors permanent and reliable dividend increases. Table 4 below indicates the number of observations in the database after and before the sampling bias elimination. Following, table 5 presents a summary statistics of the observations used in the analysis.

2002-2009 2014-2018

Active and inactive companies 1,323 1,095

Database total observations 2,213,989 2,260,754

After excluding financial sector companies 1,968,231 2,079,736 Companies after eliminating sampling bias 1,281 619 Dividends after eliminating sampling bias 15,050 4,819

Table 4.Sample observations

3.2 Methodology

This study wants to find if dividend announcements affect stock prices. Therefore, an event study will be conducted. The following subsections detail the event study methodology and the model that will be used.

2002-2009 2014-2018 Both periods

Mean Min Max Mean Min Max Mean Min Max

Cash dividend 3.314 0.004 260 2.34 0.001 200 2.912 0.001 260 Industry sector 33.54 10 60 34.43 10 60 33.98 10 60 Firm price 57.14 0.001 8950 78.68 0.0001 9950 68.05 0.0001 9950 Firm return -0.13% -100% 58.24% 1.25% -100% 209.48% 0.57% -100% 209.48% Market index price 5228 2621.12 8461.61 5497.75 3647.67 6874.67 5364.63 2621.12 8461.61 Market index return -0.0009% -11.07% 12.91% 0.03% -9.95% 6.98% 0.02% -11.07% 12.9% Table 5. Summary statistics

(23)

19 3.2.1. Event window

In order to compare the reaction of stock prices to a dividend announcement in different market phases, an event study will be conducted. Thus, the first step is to determine the event window, which will allow to analyze different days before and after the dividend announcement day. This study will use an event window that comprises five days before and five days after the announcement day (𝑇₁ = −5 and 𝑇₂ = +5). After that, this paper defined the estimation window, which is a period prior to the event window. Therefore, it has been decided to set it from 106 days to 5 days before the event window (𝑇0 = −106 and 𝑇1 = −5).

3.2.2. Model

In order to measure the impact of a dividend announcement effect, it is required to calculate the abnormal return generated over the event window. To do that, it is necessary first to calculate the normal return. Therefore, a statistical model will be used due to the advantages provided and its extensive use in other papers on the same matter. Especially, this research will employ the Single Index Market Model, which reduces biases in studies that analyze bull and bear market phases, as suggested by Klein and Rosenfeld (1987).

SIMM: 𝑅_𝑖𝑡 = 𝛼_𝑖 + 𝛽_𝑖𝑅_𝑚+ 𝜀_𝑖𝑡 𝑅_𝑖𝑡 = 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑠𝑒𝑐𝑢𝑟𝑖𝑡𝑦 𝑖 𝛼𝑖 = 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡 𝑡𝑒𝑟𝑚 𝛽_𝑖 = 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 𝑅_𝑚 = 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑡ℎ𝑒 𝑚𝑎𝑟𝑘𝑒𝑡 𝜀_𝑖𝑡 = 𝑒𝑟𝑟𝑜𝑟 𝑡𝑒𝑟𝑚

(24)

20 3.3 Results

After using the nonparametric methodology, Figure 3 presents a graph that indicates the identified bull and bear market periods. The bull phases are the yellow shaded areas, while the bear periods are the non-shaded areas. Notably, Table 6 indicates the bull and bear periods that were chosen to analyze in this study. It shows the date and duration of each market phase.

Figure 3. Bull and bear periods. Bull periods indicated by the shaded areas

Period Date Duration

Bull 1 10/2002 – 10/2007 61 months

Bull 2 07/2017 – 01/2018 19 months

Bear 1 11/2007 – 02/2009 16 months

Bear 2 05/2014 – 06/2016 26 months

Table 6. Bull and bear phases identification

Table 7 below shows the mean abnormal returns of an increase in dividend announcement for bull and bear market phases together. This study calculated the pre-event abnormal returns, which are the abnormal returns five days prior and including the event day, and the post-event abnormal returns, which are five days after and including the event day. After that, the abnormal returns are tested for significance. It can be seen that pre-event and post-event abnormal returns for both periods are positive and statistically significant at 1% level. These results are consistent with the

(25)

21

findings of Below and Johnson (1996) and confirm the validity of the signaling theory. Thus, hypothesis 1 is proved to be correct since a dividend increase announcement positively impacts stock prices shown by the abnormal returns generated.

2002-2009 2014-2018

Pre-event AR(%) 0.783 *** 0.584 *** Post- event AR(%) 0.665 *** 0.730 *** Table 7. *p < .10. **p < .05. ***p < .01

The previous table concentrated on the overall mean abnormal returns in the periods disregarding the market phases. However, the following outcomes focused on comparing the differences between bull and bear market phases. The value of the difference is calculated by the bull mean minus the bear mean, while the t-value results from the paired t-test performed. In Table 8, it can be seen that for the period 2002 to 2009, the bear phase provide significantly higher abnormal returns compared to the bull phase. The pre-event difference of abnormal returns is statistically significant at a 1% level, while the post-event difference is significant at 5% level. However, not significant results were found for 2014 to 2016.

2002-2009 2014-2018

Bull Bear Difference t-value Bull Bear Difference t-value

Pre-event AR(%) 0.660 1.084 -0.425 -3.43 *** 0.376 0.490 -0.115 -1.41 Post- event AR(%) 0.604 0.834 -0.234 -2.02 ** 0.357 0.672 -0.315 -1.23 Table 8. *p < .10. **p < .05. ***p < .01

Besides, similarly to Akron (2011), this study found a significant difference between abnormal returns in bull and bear phases on announcement day 0. Table 9 shows that the result of the difference (-0.51%) is significant at 5% for the period 2014 to 2018, but no significant evidence was found for the period 2002 to 2009. Thus, hypothesis 2 demonstrates to be right since the abnormal returns are statistically higher for bear phases than for bull phases. It is also important to note that in the days prior to the dividend announcement exist positive abnormal returns which may indicate the presence of insider trading, as mentioned by Akintade and Olade (2014).

(26)

22 2002-2009 2014-2018 Days Bull (AR%) Bear (AR%) Difference (%) t-value Bull (AR%) Bear (AR%) Difference (%) t-value -5 0.713 1.020 -0.307 -1.04 0.342 0.206 0.136 1.02 -4 0.597 0.826 -0.229 -1.16 0.354 0.321 0.033 0.19 -3 0.579 1.445 -0.866 -2.87 *** 0.635 0.570 0.065 0.3 -2 0.803 1.612 -0.809 -2.01 ** 0.288 0.855 -0.566 -2.16 ** -1 0.680 0.961 -0.281 -1.17 0.455 0.458 -0.003 -0.02 0 0.474 0.141 0.333 1 0.055 0.564 -0.509 -2.3 ** 1 0.791 0.677 0.114 0.38 0.384 0.367 0.017 0.09 2 0.616 0.777 -0.161 -0.83 0.395 0.352 0.043 0.23 3 0.622 1.353 -0.731 -1.95 ** 0.375 0.420 -0.044 -0.26 4 0.430 0.928 -0.498 -2.15 ** 0.467 0.317 0.151 0.98 5 0.630 0.784 -0.154 -0.63 0.349 1.922 -1.573 -1.24 Table 9. *p < .10. **p < .05. ***p < .01

Lastly, Table 10 shows the cumulative abnormal returns for different event windows in each bull and bear phases. The Cumulative Abnormal Returns (CAR) for the period 2002 to 2009 are significant at 1% level, thus supporting Akintade and Olade (2014), who advocate the validity of the signaling theory and the reaction of prices to dividend announcements. These results are consistent with the literature, indicating that dividends indeed covey information used for firm valuation. Also, it can be noted that the market reaction is greater in bear phases than in bull phases. Although in the period 2014 to 2018, the CAR (-1, +1) is not significant, the other windows provide statistically significant results at a 1% and 10% level. These results again provide support for hypothesis 1 since a dividend increase produces significantly positive abnormal returns, and hypothesis 2 since bear phases show higher CAR compared to bull phases. This table also reflects that the market reacts even before the announcement is made, thus corroborating the existence of insider trading (information leakage).

(27)

23

2002-2009 2014-2018

Bull (%) Bear (%) Bull (%) Bear (%)

CAR (-5, +5) 0.793 *** 1.450 *** 1.705 *** 4.990 * CAR (-4, +4) 0.642 *** 1.325 *** 1.619 *** 1.621 *** CAR (-3, +3) 0.677 *** 0.944 *** 1.238 *** 1.487 *** CAR (-2, +2) 0.524 *** 0.869 *** 0.849 *** 0.777* CAR (-1, +1) 0.334 *** 0.412 *** 0.417 0.261 Table 10. *p < .10. **p < .05. ***p < .01

4. Conclusions

This last section will provide a summary of the main important findings. After that, the limitations of this research and recommendations for future investigation will be presented.

4.1. Summary

Previous literature advocates the existence of positive and negative abnormal returns before and after a dividend announcement. Although 3 theories explain this effect, the most powerful is the signaling theory. This theory describes that dividends convey some information; thus, prices react accordingly.

However, some authors such as Below and Johnson (1996), Akron (2011), Akintade and Olade (2014), detected that the size of the stock price reaction to dividend announcements is different in bull and bear market phases. Their results indicate that the price reaction is significantly more positive for dividend increases in bear phases than bull phases. Thus, they attribute this difference to the information content of dividends. They claim that during bull phases, the reaction is smaller since investors already expect a dividend increase, while during bear phases, the reaction is stronger since a dividend provides a credible signal of future firm profitability. Finally, insider trading was implied due to the abnormal returns generated days before the dividend announcement.

(28)

24

Thus, this study performed an event study using the Single Index Market Model as a benchmark for normal returns. The results found are consistent with previous papers. The abnormal returns and cumulative abnormal returns are significant, meaning that a dividend increase announcement generates a price reaction. Moreover, it was found significantly larger abnormal returns during bear phases compared to bear phases on the announcement day. Likewise, the cumulative abnormal returns are significant for different event windows surrounding the announcement day. This indicates that the size of the price reaction by investors in different states of the economy is statistically significant for dividend increases.

After all, the findings strongly support the existence of the signaling theory. Moreover, they suggest that market phases are an essential factor to take into account. There was found significant support of the abnormal returns and cumulative abnormal returns generated around the announcement day. Also, there is evidence that proves that the size of the price effect of a dividend announcement is different in bear and bull market phases.

4.2. Limitations

One of the main limitations of this study is that it is solely focused on dividend increases. This happened due to the lack of information regarding dividend decreases in the database. This is a crucial restriction since it disallows to give a complete overview of the market reactions to different dividend changes. Likewise, due to the lack of information in the databased used, it was not possible to obtain other factors data. Since a t-test was used instead of a multiple regression model, some relevant factors that might influence the market reaction to a dividend announcement were not taken into consideration.

(29)

25 4.3. Recommendations for future research

The first recommendation is that similar research should be done in an emerging market. This would be helpful to increase the external validity of the results found in developed economies. Another recommendation regards the inclusion of earnings announcements in order to see if results are affected. Previous research has been done using this methodology, but it has not considered the differences in market phases.

5. References

Abdullah N. H., Rashid R. A. and Ibrahim Y. (2002). “The Effect of Dividend Announcements on Stock Returns for Companies Listed on the Main Board of Kuala Lumpur Stock Exchange”, Malaysian Management Journal, 6, No. 1 & 2.

Akintade,B., Ololade, F. (2014) Dividend announcement on share prices in a bull and bear market phase. Journal of Economics and International Finance, Vol 6(12), (Dec., 2014), pp. 272-283.

Akron, S. (2011). Market Reactions to Dividend Announcements Under Different Business Cycles. Emerging Markets Finance & Trade, 47, 72–85. https://doi-org.proxy.uba.uva.nl:2443/10.2753/REE1540-496X4706S505

Asquith, P., & Weiss, L. (2016). Lessons in Corporate Finance: A Case Studies Approach

to Financial Tools, Financial Policies, and Valuation. John Wiley & Sons, Inc.

https://doi.org/10.1002/9781119228899

Baker, M., & Wurgler, J. (2004). A Catering Theory of Dividends. Journal of Finance,

59(3), 1125–1165. https://doi.org/10.1111/j.1540-6261.2004.00658.x

Below S. and Johnson K. (1996), ‘An analysis of shareholder reaction to dividend announcements in bull and bear markets’. Journal of Financial and Strategic Decisions, Vol. 9,

(30)

26

Berk, J., & DeMarzo, P. (2017). Corporate finance (Fourth edition.). Pearson.

Bernheim, B., & Wantz, A. (1995). A Tax-Based Test of the Dividend Signaling Hypothesis. The American Economic Review, 85(3), 532–551.

Brown, G., & Cliff, M. (2005). Investor Sentiment and Asset Valuation. The Journal of

Business, 78(2), 405–440. https://doi.org/10.1086/427633

Denis, D., Denis, D., & Sarin, A. (1994). The Information Content of Dividend Changes: Cash Flow Signaling, Overinvestment, and Dividend Clienteles. Journal of Financial and

Quantitative Analysis, 29(4), 567–587. https://doi.org/10.2307/2331110

Edwards, S., Biscarri, J., & Pérez de Gracia, F. (2003). Stock market cycles, financial liberalization and volatility. Journal of International Money and Finance, 22(7), 925–955. https://doi.org/10.1016/j.jimonfin.2003.09.011

Ferris, S., Sen, N., & Unlu, E. (2009). An International Analysis of Dividend Payment Behavior. Journal of Business Finance & Accounting, 36(3‐4), 496–522. https://doi.org/10.1111/j.1468-5957.2009.02126.x

Frankfurter, G., Wood, B., & Wansley, J. (2003). Dividend Policy: Theory and Practice (pp. xii, 236). Academic Press. https://doi.org/10.1016/B978-0-12-266051-1.X5000-X

Gonedes, N. (1978). Corporate Signaling, External Accounting, and Capital Market Equilibrium: Evidence on Dividends, Income, and Extraordinary Items. Journal of Accounting

Research, 16(1), 26–79. https://doi.org/10.2307/2490411

Gonzalez, L., Powell, J., Shi, J., & Wilson, A. (2005). Two centuries of bull and bear market cycles. International Review of Economics and Finance, 14(4), 469–486. https://doi.org/10.1016/j.iref.2004.02.003

Hamilton, J. (1989). A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle. Econometrica, 57(2), 357–384. https://doi.org/10.2307/1912559

Harding, D., & Pagan, A. (2003). A comparison of two business cycle dating methods.

Journal of Economic Dynamics & Control, 27(9), 1681–1690.

(31)

27

Holler, J. (2012). Event-Study-Methodik und statistische Signifikanz. Edewecht: OlWIR, Verl. für Wirtschaft, Informatik und Recht.

Hu Zuguang, & Ahmed, M. (2010). Dividend Announcement Effect on Stock Return: An Event Study on Shanghai Stock Exchange. 2010 Second WRI Global Congress on Intelligent

Systems, 2, 320–324. https://doi.org/10.1109/GCIS.2010.26

Jensen, M. (1986). Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers.

The American Economic Review, 76(2), 323–329. https://doi.org/10.2307/1818789

Jin, Z. (2000). On the differential market reaction to dividend initiations. Quarterly Review

of Economics and Finance, 40(2), 263–277. https://doi.org/10.1016/S1062-9769(99)00052-6

Kato, K., & Loewenstein, U. (1995). The Ex-Dividend-Day Behavior of Stock Prices: The Case of Japan. The Review of Financial Studies, 8(3), 817-847. Retrieved June 26, 2020, from www.jstor.org/stable/2962240

Klein, A., & Rosenfeld, J. (1987). The Influence of Market Conditions on Event-Study Residuals. Journal of Financial and Quantitative Analysis, 22(3), 345–351. https://doi.org/10.2307/2330968

Kole, E., & Dijk, D. (2017). How to Identify and Forecast Bull and Bear Markets? Journal

of Applied Econometrics, 32(1), 120–139. https://doi.org/10.1002/jae.2511

Lang, L., & Litzenberger, R. (1989). Dividend announcements: Cash flow signalling vs, free cash flow hypothesis? Journal of Financial Economics, 24(1), 181–191. Retrieved from http://search.proquest.com/docview/1295936892/

Lunde, A., & Timmermann, A. (2004). Duration Dependence in Stock Prices: An Analysis of Bull and Bear Markets. Journal of Business & Economic Statistics, 22(3), 253–273. https://doi.org/10.1198/073500104000000136

MacKinlay, A.G. (1997). Event Studies in Economics and Finance. Journal of Economic

(32)

28

Mamunur R., Fauzias M., & Izani Ibrahim. (2013). Evidence of Dividend Catering Theory in Malaysia: Implications for Investor Sentiment. Contemporary Economics, 7(4), 1–138. https://doi.org/10.5709/ce.1897-9254.125

Michaely, R., Thaler, R., & Womack, K. (1995). Price reactions to dividend initiations and omissions - overreaction or drift. Journal of Finance, 50(3), 997–997.

Miller, M., & Modigliani, F. (1961). Dividend Policy, Growth, and the Valuation of Shares.

The Journal of Business, 34(4), 411–433. https://doi.org/10.1086/294442

Neves, E., Pindado J., & de la Torre, C. (2006). Dividends: New evidence on the catering theory (Vols. 2006–2014). Interuniversity Research Master and Doctorate Program (with a quality mention of ANECA) on “Business Economics”, Universities of Valladolid, Burgos, Salamanca and León (Spain). Until 2008, Interuniversity Doctorate Program (with a quality mention of ANECA) “New trends in Business Administration”, Universities of Valladolid, Burgos, and Salamanca (Spain). Master en Investigación y Programa de Doctorado Interuniversitarios (con mención de calidad de la ANECA) en “Economía de la Empresa”, Universidades de Valladolid, Burgos, Salamanca y León (España). Hasta 2008, Programa de Doctorado Interuniversitario (con mención de calidad de la ANECA) “Nuevas Tendencias en Dirección de Empresas”, Universidades de Valladolid, Burgos y Salamanca (España).

Park, K., & Jang, S. (2013). Capital structure, free cash flow, diversification and firm performance: A holistic analysis. International Journal of Hospitality Management, 33(C), 51–63. https://doi.org/10.1016/j.ijhm.2013.01.007

Pettit, R. (1972). Dividend announcements, security performance, and capital market efficiency. Journal of Finance, 27(5), 993–1007. https://doi.org/10.1111/j.1540-6261.1972.tb03018.x

Strong, N. (1992). Modelling abnormal returns: A review article. Journal of Business

Finance & Accounting, 19(4), 533–553. https://doi.org/10.1111/j.1468-5957.1992.tb00643.x

Suwanna, T. (2012). Impacts of Dividend Announcement on Stock Return. Procedia -

(33)

29

Wann C. and Lobo B. (2009), ‘State-dependent dividend preferences: the case of economic recessions’, American Journal of Business Research, Vol. 2, No. 1, pp. 21 – 37

Watts, R. (1973). The Information Content of Dividends. The Journal of Business, 46(2), 191–211. https://doi.org/10.1086/295525

Analysis of the dividend announcement effect in different market phases

Contents

1. Introduction

2. Literature Review

3. Empirical analysis

4. Conclusions

5. References