How does dividend events affect stock prices?: An event study on market efficiency

NATIONALEKONOMISKA INSTITUTIONEN Uppsala universitet

Examensarbete C

Författare: Fredrik Hansson Handledare: Lars Forsberg HT 2020

How does dividend events affect stock prices?

An event study on market efficiency

Abstract

This paper examines the effects of dividend announcements and dividend payments on OMX30 stock prices and tests if these effects indicate market efficiency. An event study methodology is used to find if the dividend events have a significant impact on stock prices. The study finds that both dividend announcements and dividend payments have a significant negative effect on prices. Disappointed investors or lowered expectations for future dividends may be the cause of the announcement effect. The results indicate that the stock market is semi-strong efficient for the announcements but inefficient when it comes to the payments.

Keywords:​ Event study, Market efficiency, Dividend announcement, Dividend payment

Contents

Introduction 3

Efficient markets 5

Event studies methodology 10

Previous studies 16

Data 19

Results 22

Conclusion 26

Sources 27

Appendix 30

Introduction

What is the relation between stock prices, dividend announcements, and dividend payments? Market reactions to dividend events are likely shaped by whether the market is efficient or not. If the market is inefficient prices on the market will not reflect the value of the stock. Moreover, If the market is efficient, it means that most forms of market analysis are a waste of time and money. If market analysis is useless then paying a fund manager is pointless since they won’t do any better than any other investor. For any investor trying to beat the market, it would be equally useless to spend time analyzing since it won’t yield any results. If markets are efficient investors should stop wasting their time trying to beat the market. For these reasons it is of interest to know if the market is efficient. (Bodie, Kane, Marcus, 2017).

Market efficiency has been discussed for a long time and is a central economic theory.

In the 1970s the market was considered to be efficient and this was widely accepted among scholars (Shiller, 2003). The 1980 saw critique being leveled against the theory for going too far in its claims and now it is considered unlikely that the market is entirely efficient. Some aspects may still be efficient though and this makes it worthwhile to explore to which degree efficiency might characterize a market.

The value of a stock should reflect the possible future earnings that it can bring to the investor. The market will value the stock by the sum of all expected future discounted dividends (Campbell, 2018) and therefore, a stock that will never give dividends does not have monetary value to the investor. An efficient market should reflect this and market prices will be decided by the prospect for future dividends (Shiller, 2003). But, what happens when new information is introduced to the market? For instance, if it is announced that the stock will give increased dividends, will this affect the price? If the announcement is unexpected and the market is efficient it might increase the price of the stock. Alternatively, in some markets, investors might prefer companies to reinvest their profit instead of paying dividends. This could be because investors expect the company to grow and in the future pay higher dividends. Paying dividends would then put off investors and the stock price would drop.

In this paper, I will explore the effect of both the announcements of increased cash dividends and the payment of those dividends on OMX30 stock prices. This will be done with market efficiency in mind. What effect do dividend announcements and dividend payments have on OMX30 stock prices and does this indicate stock market efficiency?

To answer this question, I will use an event study, a proven method that is known to be appropriate to use when testing for market efficiency (Fama, 1991). The event study methodology described by MacKinley (1997) will be used to answer the research question described above. An event study analyzes the deviation from expectations of some variable during an event. The event can be any occurrence that might affect the variable of interest. For this study, the variable of interest is stock price and the events are dividend announcements/payments. A more detailed description of the event study methodology can be found below.

This paper concludes that there is a significant negative effect on prices from both dividend announcements and dividend payments. This suggests that the market is semi-strong efficient with regard to dividend announcements since the price seems to adjust quickly to the announcement of the dividends. The significance of the dividend payments indicates an inefficient market since the dividend payments hold no

information that the investors did not already know about. Below you will find a more thorough background on the subject of market efficiency, dividend expectations, and event studies. Following this, the methodology used in this paper is described in detail, why different choices are made is explained, and the data used in the event study is described and explained. The paper continues on to describe previous studies that used the event study methodology. Lastly, results are presented and a conclusion is drawn from the results.

Efficient markets

A clear definition of market efficiency is needed to construct a testable hypothesis of it.

Most definitions use the market's relation to available information as a way to define efficiency, which Malkiel (Campbell, 2018) does with the following quote:

“A capital market is said to be efficient if it fully and correctly reflects all relevant information in determining security prices. Formally, the market is said to be efficient with respect to some information set, ϕ, if security prices would be unaffected by revealing that information to all participants. Moreover, efficiency with respect to an information set, ϕ, implies that it is impossible to make economic profits by trading on the basis of ϕ.”

Campbell (2018) writes that it is possible to construct a testable hypothesis based on the assertion made in the second sentence of Malkiel's definition. He continues, however, by describing the difficulty of actually testing the efficient market hypothesis due to the joint hypothesis problem. The problem states that it is impossible to construct tests of market efficiency except together with a model of expected returns used to measure the efficiency (Campbell, 2018). Thus, if a model is used to estimate the difference between real and expected returns and it turns out that they differ, does this mean that the model is wrong or that markets are inefficient? It is possible to alleviate the problems caused by the joint hypothesis problem by using event study methodology (Fama, 1991). Event studies find success because they can measure the impact of new information becoming available to market participants. If security prices increase (or decrease) over a short amount of time and these prices are compared to expected prices constructed with some model, the model choice becomes less important. This is because if a model used to calculate abnormal returns delivered imperfect estimates, it would be improbable that the estimations of the expected returns would deviate greatly from real returns when used to forecast only a few days ahead. If this is true, then it would seem unlikely that abnormal returns would be large during an event if the event did not have an impact on prices. If an event has an impact on prices it is possible to evaluate its effect and see if it is consistent with the efficient market hypothesis. It is

important to remember that the joint hypothesis problem never disappears completely and the more days are studied during an event the bigger the problem becomes (Fama, 1991). Event studies cannot be used to study a large number of days and still be

considered to mitigate the joint hypothesis problem.

Malkiel's second sentence of the definition of market efficiency implies that there are different kinds of information that may be relevant to security prices. Fama (1970) similarly presents the statement, saying that "A market in which prices always 'fully reflect' available information is called 'efficient.'" He then divides market efficiency into three subcategories based on which information they "fully reflect." These are the weak, the semistrong, and the strong forms of market efficiency. The weak-form includes historical data which means it includes past prices, trading volume, and short interest (Bodie, Kane, and Marcus, 2017). If a market is weak-form efficient trading securities based on historical data is meaningless because all other investors have already taken historical data into account and the prices reflect this data. You cannot beat the market by using information already incorporated into the market!

The semistrong-form includes, in addition to everything included in the weak-form, all publicly available information. Potential investors would have access to such knowledge as who the CEO of a company is, yearly reports, and if a company has any particular dividend policies. If a market is semistrong efficient, pointing out any of these facts to investors would not affect security prices. If new information were to become publicly available on a semistrong-form efficient market, prices would adjust almost instantly to the new information. Some studies show that market prices might reflect dividend or earnings announcements as quickly as ten minutes after they occur (Bodie, Kane, and Marcus, 2017).

The strong form of market efficiency states that prices reflect not only past and publicly available information but also all relevant information. All relevant information means that information that only those working at a company might know is also included (Bodie, Kane, and Marcus, 2017). Company secrets, not yet revealed dividend

announcements or annual reports, and illegal activities the company is part of would all be the kind of information that prices would need to reflect for a market to be

strong-form efficient. If the market is strong-form efficient revealing this kind of

information does not affect security prices. For example, if the market were strong-form efficient, a dividend announcement would never affect security prices since the security price would reflect the announcement as soon as the company decided on how large the dividend should be.

Shiller (2003) describes the market as efficient when the price of a stock is equal to the present value of all future dividends conditional on the information available at the time.

However, it is not possible to know the future dividends, so a market is efficient if the stock price is equal to the optimal forecast of the present value of future dividends.

Equation 1 P P_t = E_{t t}^*

Equation 1 shows how the price of a security (P) depends on all the expected (E) future dividend payments (P*) at time t

According to Shiller (2003), any deviation in price from the model must be due to new information that makes investors reconsider the value of future dividends. Since P* is the present​ ​value of future dividends, the forecasts of future dividends need to be discounted. The discount rate can differ between models but Campbell and Shiller developed a model for asset pricing that is suitable to use as an example (Campbell, 2018). The model is an approximation but illustrates the relation between price and dividends well. Their approach starts from the definition of returns as the sum of price and dividend divided by last period's price as seen in log form in equation 2.

Equation 2

og(P ) og(P )

r_t+1 = l _t+1 + D_t+1 − l _t

Equation 2 shows how the log return (r) in period t+1 depends on the price (P) of the security and the dividends the security pays (D)

Equation 3 shows that prices depend on future discounted dividends and that if future expected dividends increase, then the price will also increase. If a surprisingly large increase in dividends were to be announced, then, as long as it did not lower the

expectations on future dividends, one would expect the price to go up as a result of the announcement. If, however, a company announced a lower then expected dividend, a reduction in the price is to be expected instead. If a dividend were announced and it did not affect the price, then, according to the model, this dividend was the one that the market expected.

Equation 3

p_t = _1−ρ^k + p_{CF ,t}+ p_DR,t

Equation 3 shows how the log price (p) of a security in time period t depends on the constants k and ρ ​as well as the price driven by expected cash flows and the price driven by the expected discount rate

Equation 4

(1 )d p_{CF ,t} = E_t ∑^∞

j=0ρ ^j − ρ _t+1+j

Equation 4 shows the price driven by expected cash flows where ρ ​is a constant and d are future dividends

Equation 5

− r

p_DR,t = E_t ∑^∞

j=0ρ^j _t+1+j

Equation 5 shows the expected future discount rate where r_t is log stock return for time period t.

It is also assumed that equation 6 holds.

Equation 6 p

j→∞limρ^j _t+j = 0

Equation 6 shows that as j approaches infinity the product of to the power of j and the log ofρ

the stock price (p) j time periods ahead is 0

If announcements of dividend increases and dividend payment increases are examined using the theory above one would have certain expectations about the results of the event study. Starting with the dividend payments, there is no reason to expect prices to adjust on the payment date. The dividend payment date does not carry any new

information to investors since the dividend has already been announced. On the

dividend payment day, investors are already locked in to receive the dividend since the ex-dividend date occurs before the payment day. The ex-dividend date is the first day the stock starts trading without paying dividends. Leading up to the ex-dividend date the price remains high since investors know they will receive dividends if they own it before the ex-dividend date. When the ex-dividend date arrives the stock price drops in value equal to the size of the dividend. Since this effect has already occurred when the dividend payment day comes, and the day itself carries no additional information, the stock price should not change. If a significant change occurs, it is either due to the market being inefficient or due to another event causing the price to shift.

Dividend announcements, on the other hand, provide investors with new information.

The stock price should adjust to reflect this new information. According to Fama (Fama, 1991), the adjustment should take place within a day. A quick price adjustment indicates that the market is semi-strong efficient with regards to dividend announcements.

However, if the market does not adjust, it does not necessarily mean that it is inefficient.

The dividend's size could be the one investors were expecting and then the price would not need to be adjusted, even if the market is efficient. Lastly, it could be that increased (decreased) dividends today are canceled out by expectations of lower (higher)

dividends tomorrow, causing the prices to remain the same.

Event studies methodology

To find out whether dividend announcements or payments affect prices, this paper uses an event studies method as described by MacKinley (MacKinley, 1997). The idea

behind this method is to define an event, for example, a dividend announcement, and try to find out if a variable of interest during the event deviates from the norm. If the deviation is statistically significant the event has affected the price. Below I will outline the process of an event study and describe the choices made during the study.

To start, it is necessary to define the event that will be studied. In this paper, it will be dividend announcements and dividend payments. It is easy to define the day on which the event occurs but, the event often affects the price of a stock during more than just the event. To take this into account an event study usually looks at several days around the event. This event study looks at one day before the event and five days after the event in addition to the event day itself. This seven day period of interest is called an event window. The choice of the size of the event window varies depending on the event studied. Previous studies have shown that the market usually reacts fairly quickly to dividend announcements, sometimes as fast as 10 minutes (Bodie, Kane, and

Marcus, 2017) and therefore a larger window is probably redundant. It is also important to make sure that no other event occurs during the event window. If another event happens during the event window that has an effect on the variable being studied it might produce a biased estimate of the effect the event has. In the case of this study, ex-dividend dates are known to affect asset prices (Campbell and Beranek, 1955) and they often occur close before the dividend payment, making it important to make sure they are not included. For this reason, the event window is limited to only incorporate one day before the event so as not to include the ex-dividend date.

It is not possible to study all dividend announcements and all dividend payments.

Therefore, this paper limits itself to study the OMX30 stock dividend events that saw increased dividends during the 2010s. Increased dividends are used because they are more prevalent than dividend decreases and mixing increases and decreases might make the effects on prices cancel out each other due to how the event study

methodology is carried out. From the OMX30 stocks, some dividend events are chosen to conduct the study. More on the selection of event dates and stocks are found below.

The event study works by comparing the stock returns that are observed during the event to those that would be expected had the event not occurred. Since it is impossible to observe what would have happened during the event windows if the events did not occur this paper forecasts the expected normal returns during the event windows. To create a forecast, data from before the event window is needed to construct a model.

The data consists of prices taken from trading days prior to the event window and is called an estimation window. The estimation window is usually much larger than the event window since more data helps to make more accurate forecasts. A 120-day estimation window is constructed for the dividend announcements and a 200-day estimation window is used for the dividend payments. Both of the windows end the day before the event window begins. The reason that the two event windows are of different sizes is that it is desired to avoid other events occurring during the estimation window.

Specifically, some stocks have bi-annual dividends and if the estimation window is too long it might include the previous dividend events making the estimation of normal returns biased. Since the dividend payments occur after the dividend announcements the payments have the option of a longer estimation window. If too many other events are included in the estimation window, it may skew the forecasts.

Figure 1 - Event study graphics by Meyer, Gremler, and Hogreve (2014)

Figure 1 shows how the estimation window and event window are constructed and what steps of the event study methodology is conducted in each window

Making the forecasts requires creating a model using data from the estimation window.

Two common models used in event studies are the constant mean return model and the market model. The constant mean return model is a fairly simple model that assumes that the returns of stocks are equal to some mean value plus a normally distributed error term. This model simply calculates the mean return during the estimation window and assumes that the same pattern will continue during the event window.

Equation 7 R_it = μ_i+ ζ_it

Equation 7 shows the constant mean return model where the return (R) for event i and time t are estimated by the mean return (μ) ​of the estimation window plus some error term ( )ξ

The second method, the market model, regresses the stock returns on the returns of a market index. The method produces a regression model that can then be used during the event window together with market index data to estimate normal returns during the event window. Returns are used instead of prices and market index in the model so that problems with autocorrelation does not affect the results. If a time-series with

autocorrelation was used to estimate the market model it might cause the model to over or underestimate how closely the stock follows the market index. Since returns are the first difference between prices this negates the problem of autocorrelation.

Equation 8

) )] 00

R_ti = [ln(P_ti − ln(P_t−1,i * 1

Equation 8 shows how returns are calculated in this study. The returns of event i at time t depends on the price (P)

Equation 9 R R_it = α_i+ β_{i mt} + ε_it

Equation 9 shows the market model where the estimated return (R) for event i at time t is estimated by an OLS regression with the market return as independent variable

Since the market model takes the movement of the market into account it will be used to conduct the event study. The mean return model risks missing important reasons for why the stock varies in value due to being too simple. It is possible to construct more complex models to measure normal returns during the event period that might be even better, but that is beyond the scope of this paper. For this reason, the market model will be used to estimate normal returns.

Equation 10 AR

︿

= R_iτ − α^︿_i− β R^︿_{i mτ}

Equation 10 show how abnormal returns (AR) are equal to the difference between the actual returns and the market model estimated returns for event i at time τ ​which is event time

Equation 11

CAR

︿

= ∑

T_j

τ =T_n+1AR

︿

Equation 11 show how the cumulative abnormal returns (CAR) are equal to the sum of all abnormal returns during an event window for event i and event time where τ ^T_n+1 ​and​ ^T_j ​are the first and last days of the event window

The event study method uses abnormal returns (AR) as described in equation 10 to measure if the returns during the event deviate from what would be expected. In this paper, abnormal returns are the difference between the market model estimated normal returns for the event window and the actual returns observed during the event window.

For each day of the event window, normal returns are compared to observed returns, and the difference between the returns are calculated. These AR can then be summed over the event window, resulting in the cumulative abnormal returns (CAR) of the event.

This is done for every event of interest and the CARs of the events are tested to see if they differ from zero. If they do, we can conclude that the event has had an impact on

prices. The test of the statistical significance of the CARs is made using a t-test. The null hypothesis is that the CARs are zero since this would mean that they do not deviate from the normal returns that are expected to occur. The alternative hypothesis is then that the CARs are not zero and that the event has had an effect on prices. The ARs are assumed to normally distributed with mean zero under the null hypothesis since only chance or measurement errors would make them deviate from the expected normal returns. Following this, the CARs are also normally distributed with mean zero and some variance since the sum of two normally distributed variables are also normally distributed (Hogg and Tanis, 1997). When testing the significance of the CARs a student's t-test is, therefore, appropriate since a test of significance is made on the sample drawn from a normally distributed variable.

Equation 12

σ²_εi = _L¹ (R )

1 ∑^Tn

τ =T₁ iτ − α^︿_i− β R^︿_{i mτ} ²

The disturbance variance σ² ​is equal to the summed squared errors of the market model during the event window divided by the length of the estimation window L₁

Equation 13

(AR ) [1 ]

σ²

︿

= σ²_εi + _L¹

1 +

︿2σ

m

(R −μ )_mτ ^︿_m²

The variance of the abnormal return estimate is equal to the disturbance variance from equation 12 plus a second term where L₁ ​is the length of the estimation window, R_mτ ​is the market return during the event window day τ , ​^{and ︿}μ_m ^{and ︿2}σ_m is the mean return and return variance of the market during the estimation window. If L₁ ​ is large the second term of the equation can be ignored

Equation 14

) (AR )

σ (CAR²

︿

i = ∑

T_j

τ =T_n+1σ²

︿

The variance of the CAR estimate for event i is equal to the sum of the AR for event i over the event window starting with day T_n+1 ​and ending at day T_j

Equation 15

ar(CAR)

v = _N¹2 ∑^N

i=1 σ (CAR²

︿

i

The variance of the average CAR is equal to the sum of the CAR variances for events 1 through N divided by N squared where N is the number of events in the event study

Equation 16

t =

Equation 16 shows that the t-statistic is equal to the average CAR for all the events divided by the root of the variance from equation 15

Now that a t-test has been made, it is possible to accept or discard the alternative hypothesis and go on to analyze the results.

Previous studies

There have been many attempts to measure the effect that dividends have on stock prices. Below are five examples of different papers testing the effect that dividends have on prices. These five papers do not come to the same conclusion, some find a

significant effect on prices and others do not.

Ali and Chowdhury (2010) analyze the effect of dividend announcements on the prices of the banking industry in Bangladesh. They use an event study methodology to find results. The sample used consists of 25 Bangladeshi banks that announced dividends between January 1st or a maximum of 30 days later. They use the event study

methodology described by MacKinley (1997) as a jumping-off point, but they fear the bias that their results might become biased due to how frequently the banking stock is traded and therefore modify the methodology. They adjust the returns of the banking stock based on the average growth of the industry and then they continue with the event study methodology. Their results show that the dividend announcements have no

significant effect on stock price and they speculate that this might be due to insider trading, making the dividend announcement an event that provides no new information.

The second paper, by Udim and Udim (2014), also studies the effect of dividend announcements on the banking industry in Bangladesh. They too use an event study method, but it is different from the study by Ali and Chowdhury since they stick closer to the methodology as it is described by MacKinley (1997). Using data on bank stocks traded on the Dhaka stock exchange and a Dhaka stock exchange index they construct a model for normal expected returns. Their results show that there is no significant difference between the expected price and the actual price. This means that they find no evidence that dividend announcements affect the stock price.

Dinh and Nguyen (2016) also studied dividend effects on prices but, in addition to studying the effect of dividend announcements, they also try to measure the effect the ex-dividend date has. Their data is extensive, using 1962 dividend events from

Vietnamese companies between the years 2008 and 2015. To find the effect of both the

ex-dividend date and the dividend announcement date they use the methodology similar to that described by McKinley. Their results show that the effect on stock price from a dividend announcement is positive and that the effect on price from the ex-dividend date is negative. They speculate that the positive effect dividend announcements have on prices might come from investors following the bird-in-hand theorem which states that investors prefer dividends over future capital gains since capital gains are less reliable than dividends. This would make the announcements of dividends something that investors prefer and therefore prices would increase.

Kadıoğlu, Telçeken, and Öcal (2015) studied the effect of dividend announcements on Turkish stocks. They studied 902 dividend announcements between 2003 and 2015.

They too measure the difference between an estimated normal price during the dividend announcement event and the actual observed price. By using the Turkish market index, they construct a model using a market-adjusted model, which compares the return of the market to the return of the stock without first estimating a model. This differs from both Udim and Udim and Dinh and Nguyen who both use regression models to estimate their normal returns. They then use dividend per share to regress on abnormal returns to find if increased dividends per share affect abnormal returns. The authors conclude that there is a statistically significant negative effect on prices when dividends are announced. They believe it may be due to Turkish tax laws since capital gains are not taxed as hard as cash dividends in Turkey which would make the dividends less desirable causing investors to sell their stock to buy something that generates greater capital gains.

Michaely, Thaler, and Womack (1994) wrote a paper studying how the announcement of dividend initiations and omissions affect excess dividends. They study stocks traded on the NYSE and the AMEX during the years 1964 through 1988. Among other event windows, they look at the day before, the day off, and the day after the

initiation/omission events. Their sample consists of 235 initiation and 290 omission events. To find the excess returns they use stocks similar to the stocks they wish to study. These similar stocks are stocks of companies that are of the same size as the

studied stocks. The difference in return between the studied and comparison stocks during the event window is the excess return of the studied stock. They find that the return during dividend initiations is 3,4% and the return during dividend omissions is -7,0%. These results would indicate that investors (during the years 1964 through 1988 and for the NYSE and AMEX stock markets) like dividends and dislike omitted

dividends.

Data

The event study will be conducted on OMX30 stock price data to see how they are affected by the dividend events. The price data used in the study come from Nasdaq nordic which is a subsidiary of Nasdaq, Inc (Nasdaq nordic, webpage). Nasdaq, Inc is a well-known company that delivers many financial services. The data from Nasdaq was downloaded as several time-series, one for each stock studied. Each time-series contains daily data on stock prices but non-trading days aren't included in the dataset.

Some holidays are included in the data and take on the value of the last trading day.

These days have been manually removed from the data since they are not trading days and hold no information. All of the time-series start on 2010-10-11 and end on

2020-10-09.

Woodseer has been used to find out on which dates dividend announcements occurred, what the size of the dividend increase was, and when the dividends were paid.

Woodseer is a company that specializes in making dividend forecasts. 26 dividend announcements and 26 dividend payments have been chosen as events of interest.

These events have been chosen based on data availability and the dates of dividend announcements were the hardest to come by and they, therefore, limited the available data the most. This has resulted in 16 different OMX30 stocks being studied. The chosen stocks as well as the dividend announcement and payment dates can be seen in table 1.

Since the study uses the market model to calculate normal returns during the event windows, a proxy for the market is needed. The OMX Stockholm Price Index (OMXSPI) will be used as a substitute for the market. Nasdaq nordic has once again been used to retrieve this data. The OMXSPI is a price index for all the stock traded on the Stockholm stock exchange and is a reasonable proxy for the market index since it is exposed to many of the same market events as the OMX30 stocks. This is because they are both traded in the same markets. This paper will use closing prices for both the OMX30 stocks and the OMXSPI. A stock's last trade may not occur at the same time as another and this may cause problems if an event happens between the two last trades since it

may skew the difference between the two stocks. This is probably a minor problem for the OMX30 stocks because they are frequently traded and it is far better to use the closing price than the highest or lowest price during the day. If the highest or lowest price is used instead there is no way of knowing at which time of the day the trade happened and if it captures an event or not. This may make it so that the analysis misses the effect of the event.

Table 1 - Dates for dividend announcements and dividend payments

OMX30 Stocks Dividend announcement date (Y-M-D) Dividend payment date (Y-M-D)

ABB 18-02-28 18-05-15

Alfa Laval 19-02-05 19-05-02

ASSA ABLOY 17-02-02 17-05-04

19-03-19 19-05-03

Astra zeneca 12-02-02 12-03-19

12-07-26 12-09-10

Atlas copco A 12-02-01 12-05-08

13-02-01 13-05-08

Atlas copco B 11-02-08 11-05-02

12-02-01 12-05-08

13-02-01 13-05-08

Boliden 17-02-10 17-05-03

19-03-27 19-05-10

Ericsson B 15-01-27 15-04-21

16-01-27 16-04-20

Handelsbanken A 13-02-07 13-03-28

15-02-04 15-04-01

Hexagon B 19-02-28 19-04-17

Securitas B 18-01-31 18-05-09

Skanska B 14-02-07 14-04-11

15-02-13 15-04-16

SSAB A 19-02-28 19-04-15

Swedish match 17-02-20 17-05-11

Telia 11-02-03 11-04-14

14-01-30 14-04-10

Volvo B 17-02-01 17-04-11

Results

The two tables in the appendix (table 3 and table 4) show the result for the market model estimates. Every model consists of an intercept and a slope coefficient as well as the regressor and the regressand as described in equation 9. None of the intercept coefficients are statistically significant but all of the slope coefficients are. The intercept holds no information of interest, since the OMXSPI is unlikely to fall to zero, and

therefore its lack of significance is not a problem. It is more important that the model as a whole has significance, otherwise one might suspect that it is not a good model to estimate the normal return with. An F-test is used to test model significance and all of the models have a p-value that is close to zero meaning that the OMXSPI returns provide useful information to predict the stock returns. The significance of the slope coefficients also confirms this. The F-tests for the dividend announcements have been made with 1 and 118 degrees of freedom and the F-tests for the dividend payments have been made with 1 and 198 degrees of freedom.

According to table 2, the t-test of the CARs are significantly different from zero for the dividend announcements and the dividend payments. If the returns differ from zero, the price change during the event window also differs from zero. A significant price change during the event window suggests that the market has reacted quickly to the

announcement. The quick reaction indicates that the dividend announcement was unexpected but that the market incorporates the information into the stock price efficiently. The market, therefore, seems to be semi-strong efficient since it adjusts to reflect new prices. Since the CARs are negative, the price drops during the event window. Even though the dividends increase, investors lose faith in the stock. There are several possible explanations as to why this is.

Table 2 - Event study results

P-value for the dividend announcements

P-value for the dividend payments

L_1 is considered small 0,03147 0,001003

L_1 is considered large 0,03264 0,001039

Table 2 shows the p-values for the significance test of the CARs from equation 16. Significance means that the CARs deviate from the null hypothesis that they are zero. The deviation is negative for both the announcements and payments and the t-tests are two-sided and have been made with 25 degrees of freedom.

One possibility is that investors expected the dividends to be bigger. The price of the stock might have reflected expectations of a larger dividend. Investors would then be disappointed by the dividend announcement since their expectations are not met. The price would then go down when the investors are disappointed by the size of the dividend. Overly optimistic expectations of the size of the dividend might be due to the strong bull market during the 2010s. If the market is strong, the investors might expect the stock to pay large dividends based on previous dividend-price ratios. A lower dividend-price ratio could cause investors to find the stock less valuable. Because the investors seem surprised by the dividend announcement, it indicates that the market is not strong form efficient. If the market was strong form efficient, investors could never be surprised by a dividend announcement. They would already know how large the dividend would be because of insider information. The price would then already reflect the dividend and the market would not be surprised by the announcement.

Another possibility is that investors thought the dividend was too large, not too small.

Investors might suspect that a large dividend now might mean smaller payments in the future. If the dividend is so large that it is improbable that the company will continue to pay dividends that large, investors will then consider the stock less valuable. According to equations 3 and 4, the stock price depends on the discounted value of all future dividends. Even if the next dividend is large, the price may still go down if the future expected dividends declines. In a similar vein, investors might not like that the company pays dividends at all. Investors might want the company to reinvest all of their profits. If the investors believe that the highest return they can receive comes from reinvesting in the company, they would dislike receiving dividends. Dividends now would, once again, mean lower dividends tomorrow, causing the price to drop.

One last possible explanation is that the cost of getting dividends are higher than those of capital gains. Kadıoğlu, Telçeken, and Öcal (2015) reasoned like this when they received a similar result as in this paper. They claimed that the lower price in association with the dividend announcements was due to Turkish tax laws where dividends are taxed more than capital gains. This is not the case in Sweden, where capital gains and dividends are taxed the same for OMX30 companies.However, there might be some other hidden cost that causes investors to be averse to dividends, which could cause the observed price drop.

The returns during the dividend payment event windows are also significantly different from zero. Since the returns are negative, the price has dropped during the payment event window. The payment of a dividend, as previously explained, should not affect the price of a stock on an efficient market because it does not provide investors with new information. As to why the market still reacts, one could think of a couple of reasons.

Firstly, it could be a residual effect from the ex-dividend day. A stock dropping in price on the ex-dividend day is expected. On an efficient market, the price drop should be contained to the ex-dividend day because investors should know that they are no longer entitled to dividends when they trade the stock. However, if some investors are

inattentive and only realize that the ex-dividend day occurred some days after, then they may sell their stock during the event window, causing the price to drop.

Secondly, the same thing could happen if they are not aware of how dividend payment works or if they think that the stock is decreasing in value for some other reason. In that case, they might hold on to the security until the dividend has been paid and only

afterward sell it, falsely believing that they must own the stock when the dividend is paid. If an investor, somehow, was not aware that the ex dividend day had occurred and just watched the price of their stock plummet, they might sell it off believing it might become worthless, causing the price to drop even after the ex dividend day. Lastly, the result might just be due to an unlikely outcome that could have occurred even if the confidence level is very high. It is unclear as to why arbitrageurs are not taking

advantage of this effect. Knowing that a stock will decrease in value when a dividend is

paid would make it easy to short the stock just before the dividend payment. One possible reason is that this is a new phenomenon and has not been a reliable arbitrage opportunity before. Investors would then not know about it as a possibility to earn money.

Conclusion

This paper has explored the effect that increased cash dividend announcements and payments have on security prices. To answer the research question an event study was used. It utilized OMX30 and OMXSPI stock data to estimate abnormal returns during seven days around the dividend announcement and payment dates. For 25 different stocks, the cumulative abnormal return was calculated for each stock, and a t-test was used to test if the CAR were significantly different from zero. The test shows that the CARs are significant and negative for both the dividend announcements and the dividend payments. It means that the stock price decreased during both events.

These results may seem surprising. It is a bit counter intuitive that prices would

decrease when dividends are raised. The quick decrease in price is however consistent with a semi-strong market and could be explained either by investors being

disappointed by the size of the dividend or that they expect lower future dividends. On the other hand, the dividend payment result does not suggest that the market is

efficient. The dividend payment event does not contain any new information for

investors and should therefore not affect the price. If the market is inefficient there might be possibilities for arbitrageurs to make money with little risk. However, further study would probably be prudent before staking your life savings on this arbitrage opportunity.

The results could also be strengthened by more data. A relatively short time was possible to use for data gathering and having data on not only 16 OMX30 companies but also the rest of them could improve results. Studying more dividend events could also grant greater insight. I have only studied dividend increases in this paper, but it is possible to also use decreased dividends or stable dividends to compare. Maybe the results look different when dividends are not increasing?

Sources

Hogg, R. Tanis, E. (1997), ​Probability and statistical inference​, Prentice Hall, New Jersey.

Campbell, J. (2018), ​Financial Decisions and Markets​, Princeton University Press, Oxfordshire.

Bodie, Z. Kane, A. Marcus, A. (2017), ​Essentials of Investments​, McGraw-Hill Education, New York.

Shiller, R. (2003), “From Efficient Markets Theory To Behavioral Finance”, ​Cowles Foundation Paper​ No. 1055.

Ali, M. Chowdhury, T. (2010), “Effect of Dividend on Stock Price in Emerging Stock Market: A Study on the Listed Private Commercial Banks in DSE”, ​International Journal of Economics and Finance​, Vol 2, No. 4, 52-64.

Uddim, N. Uddim, M. (2014) , “Dividend Announcement of the Commercial Banks in DSE: Scenario and Effect on Stock Price”, ​International Journal of Ethics in Social Sciences​, Vol 2, No. 1, 143-155.

Dinh, B. Nguyen, C. (2016), “Dividend Announcement and Ex-Dividend Effects on Stock Return”, ​International Journal of Economics and Finance​, Vol 8, No. 7, 207-215.

Kadıoğlu, E. Telçeken, N. Öcal, N. (2015), “Market Reaction to Dividend

Announcement: Evidence from Turkish Stock Market”, ​International Business Research​, Vol 8, No. 9, 83-94.

Samitas, A. Kenourgios, D. (2004), “Market Efficiency and Signaling: An Event Study Analysis for Athens Stock Exchange”, ​Proceedings of the 1st Applied Financial

Economics (AFE) International Conference on “Advances in Applied Financial Economics”, pp. 163-175, Samos island, Greece​.

Malkiel, B. (2005), “Reflections on the Efficient Market Hypothesis: 30 Years Later”, ​The Financial Review, 40​, 1-9.

Fama, E. (1970), “Efficient Capital Markets: A Review of Theory and Empirical Work”, The Journal of Finance, ​Vol 25, No. 2, 383-417.

Fama, E. (1991), “Efficient Capital Markets: II”, ​The Journal of Finance, ​Vol 46, No. 5, 1575-1617.

McKinley, A. (1997), “Event Studies in Economics and Finance”, ​Journal of Economic Literature​, Vol 35, 13-39.

Meyer, J. Gremler, D. Hogreve, J. (2014), “Do Service Guarantees Guarantee Greater Market Value”, ​Journal of Service Research​, Vol 17, 150-163.

Michaely, R. Thaler, R. Womack, K. (1994), “​Price Reactions to Dividend Initiations and Omissions: Overreaction or Drift?​”, National Bureau of Economic Research, Working paper No. 4778.

Campbell, J. Beranek, W. (1955), “​Stock Price Behavior on Ex-Dividend Dates​”, The Journal of Finance, Vol 10, No. 4.

Nasdaq Nordic. “Historiska kurser - Aktier”, (2020-11-10).

http://www.nasdaqomxnordic.com/aktier/historiskakurser

Nasdaq Nordic. “Historiska kurser - Index - OMXSPI, OMX Stockholm_PI”, (2020-11-12).

http://www.nasdaqomxnordic.com/index/historiska_kurser?Instrument=SE0000744195

Woodseer. “Nasdaq Nordic Stockholm XSTO”, (2020-10-09).

https://www.woodseer.global/stock_exchanges/nasdaq-nordic-stockholm-XSTO/

Appendix

Table 2 - Market model estimates for dividend announcements

OMX30 Stocks Intercept Slope F-statistic p-value

ABB 0.0003933 1.1323571*** 0,000

Alfa Laval -0.0007311 1.2689883*** 0,000

ASSA ABLOY 2017 and 2019

-0.001604 1.089119*** 0,000

0.0009647 0.8392761*** 0,000

Astra zeneca 2012 feb and july

0.0007889 0.4062290*** 0,000

1.189e-05 0,3027*** 0,000

Atlas copco A 2012 and 2013

0.0008422 1.2345032*** 0,000

0.0004015 1.5070915*** 0,000

Atlas copco B 2011 , 2012 and 2013

0.0011711 1.4470627*** 0,000

0.0008422 1.2345032*** 0,000

0.0004534 1.5794880*** 0,000

Boliden 2017 and 2019

0.001777 1.071420*** 0,000

0.0008228 1.4986051*** 0,000

Ericsson B 2015 and 2016

0.0005170 0.9373534*** 0,000

9.695e-05 1.146*** 0,000

Handelsbanken A 2013 och 2015

-3.213e-05 0,9251*** 0,000

2.542e-05 0,9857*** 0,000

Hexagon B -0.0003805 1.4744310*** 0,000

Securitas B 0.0007358 0.0017920*** 0,000

Skanska B 2014 och 2015

-9.378e-05 0,944*** 0,000

0.0010254 1.1115569*** 0,000

Table 2 shows the market model estimates for the dividend announcements. P-values: 0 ‘***’, 0,001 ‘**’, 0,01 ‘*’, 0,05’.’, >0,1 ‘ ‘

Table 3 - Market model estimates for the dividend payments

SSAB A -0.0001859 1.5426763*** 0,000

Swedish match -0.0010419 0.6597292*** 0,000

Telia 2011 and 2014 -0.0005324 0.5410876*** 0,000

0.0001927 0.6988570*** 0,000

Volvo B 0.0012146 1.1064736*** 0,000

OMX30 Stocks Alfa Beta F-statistic p-value

ABB 0.0002262 1.1785367*** 0,000

Alfa Laval -0.0001836 1.2087759*** 0,000

ASSA ABLOY 2017 and 2019

-0.0004606 1.0463183*** 0,000

0.0003392 0.7955928*** 0,000

Astra zeneca 2012 mar and sep

-0.0001136 0.3927790*** 0,000

-2.913e-05 0,3253*** 0,000

Atlas copco A 2012 and 2013

0.0003203 1.2143274*** 0,000

-0.0004107 1.4647014*** 0,000

Atlas copco B 2011 , 2012 and 2013

0.0008645 1.5472008*** 0,000

0.0002926 1.2667109*** 0,000

-0.0004488 1.5158219*** 0,000

Boliden 2017 and 2019

0.0009025 1.1489123*** 0,000

-0.0006567 1.6243767*** 0,000

Ericsson B 2015 and 2016

0.0004972 0.9957518*** 0,000

-0.0002613 1.0969664*** 0,000

Table 3 shows the market model estimates for the dividend payments. P-values: 0 ‘***’, 0,001

‘**’, 0,01 ‘*’, 0,05’.’, >0,1 ‘ ‘

Handelsbanken A 2013 och 2015

0.0004839 1.0129801*** 0,000

-3.826e-05 0,9697*** 0,000

Hexagon B -0.0002531 1.3985231*** 0,000

Securitas B 5.133e-05 0,9981*** 0,000

Skanska B 2014 och 2015

0.0004801 0.9489041*** 0,000

-8.847e-05 1.146*** 0,000

SSAB A -0.001235 1.650022*** 0,000

Swedish match -0.0007599 0.5708643*** 0,000

Telia 2011 and 2014 -0.0004660 0.6531194*** 0,000

-0.0005154 0.7666858*** 0,000

Volvo B 0.0010387 1.2192370*** 0,000