Seasoned equity offer issue method: does it matter?
Abstract
Research done by Burton, Lonie & Power (1999) suggests that some types of seasoned equity offer issue methods are more costly than others. In their research, they find a 7% gap between the negative abnormal returns of rights and non-rights issues, with rights issues being the worse of the two. Others find the exact opposite, stating that rights offerings have a less negative abnormal return than non-rights offerings(Eckbo & Masulis, 1992)(Armitage, 1998). Is there a difference in market reaction when a firm decides to do one or the other? The goal of this thesis is to answer the question whether or not there is a relationship between issue method and abnormal returns. The findings in Burton, Lonie & Power (1999) hold, although with different numbers than the ones presented in their research. A
significant -0.8% cumulative average abnormal return is found for rights offerings, as opposed to a 0.3% cumulative average abnormal return for non-rights offerings.
Name: Jordy Pieterman Student number: 10460438
Specialization: Finance & Organization Field: Finance
2
Verklaring eigen werk
Hierbij verklaar ik, Jordy Pieterman, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.
Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.
4 1 Introduction
Theory first presented by Modigliani & Miller (1958) introduced the world to the idea that capital structure should not matter in corporate financing. Despite the fact that this would only be the case in perfect capital markets, a starting point for future theories was made. Later theories suggest that debt is chosen such that there is a perfect balance between the cost of financial distress and the interest on the tax shield(Kraus and Litzenberger, 1973). Theoretically speaking, the value of a company should not change whether a company chooses to issue equity or fund from loans. However, empirical research proves otherwise (Marsh, 1982. Bhandari, 1988). In general, an equity issue generates a negative reaction and a debt issue does not (Korwar & Masulis, 1986). The main accepted explanation behind this phenomenon is that equity issues are viewed as a lender of last resort . When companies can’t borrow from banks, they are ought to be in financial distress or their investment opportunities are not profitable (Burton & Power, 2003). An equity issue can be offered in a variety of ways. Burton, Lonie & Power (1999) divided equity issues between two groups of offering method, rights and non-rights offerings, and analyzed differences between these groups. A rights offering is an equity issue that is first offered to existing shareholders, non-rights offerings are all other types of offerings. They found that the average 3% negative abnormal return caused by seasoned equity offerings in the United Kingdom was mainly a result of the 8% negative abnormal return of rights offerings, against only 1% insignificant negative abnormal return of non-rights offerings (Burton, Lonie & Power, 1999). Other papers report the opposite, stating rights issues have a more positive effect than non-rights issues (Eckbo and Masulis, 1992. Bohren et al. , 1997). Which of these conclusions hold during the crisis in Western Europe? I am trying to find out whether or not this ‘issue effect’ when a seasoned equity offer is issued still exists and, if it does exist, which type of offering has a more negative effect?
The main finding I will clarify in this thesis is whether or not a rights-offering in Western Europe during the crisis has an abnormal return different compared to a non-rights offering. To research the effect of issue method on market reaction, the following research question was stated: ‘’Do rights issues have a different market reaction than non-rights issues during
5
the crisis? The division of all methods into two groups has three reasons. First, comparing results to the last paper written about this subject will be more convenient. Second, the division between the rights and non-rights issue method compiles shares in groups where companies decide whom the shares go to. As will be made clear in the literature review, some companies might value existing shareholders over new shareholders (Myers and Majluf, 1984). If this is true, the company will make rights offerings when stock is
undervalued and non-rights offerings when stock is overvalued. This statement alone could possibly explain why some companies offer shares to new shareholders and some to existing ones. The last reason for the division in these two groups lies in the fact that there are too many methods of issuance compared to actual available data. For some offerings, only a handful of obtainable observations is available.
2 Literature review
A wide variety of literature exists regarding abnormal returns and seasoned equity offers. However the specific topic of this thesis, the ‘issue effect’, is something researchers have not commonly written about. A variety of articles was used to write this thesis, but Burton, Lonie & Power’s (1999) article together with Mackinlay’s paper (1997) regarding the market model are the main guidelines used to write the empirical part of this thesis. In the following paragraph the most important literature used for this thesis will be discussed.
2.1 literature
Among others, existing literature by Marsh (1982), Asquith & Mullins (1986) and Schipper & Smith (1986) predict a negative abnormal return for seasoned equity offerings. Burton, Lonie & Power (1999) investigated whether or not this statement holds for different types of issue method. Their first test was purely on issue method alone, dividing issue method between rights and non-rights issues. This resulted in a non-significant abnormal return for non-rights offerings of -1%, and a significant abnormal return of -8% for rights offerings. However, the mean shares issued per offering differed across the two issue methods, with
6
rights issues being three times larger than non-rights issues on average. A second
regression including the variable issue size was added to the regression, which resulted in a significant beta for the rights issues and a non-significant effect of issue size, suggesting that issue method is the main determinant for the market reaction on seasoned equity offerings.
Armitage (1998) reviews some evidence on rights issues and seasoned equity offerings in his paper, as well as differences between issuing process in the United Kingdom and the United States. ‘‘The pre-emptive right of first refusal is a long tradition and is part of the London Stock Exchange's listing requirements’’ (Armitage, 1998 ). All equity issues have to be
offered to existing shareholders first, before they can be sold to new shareholders. Burton & Power (2003) investigate the motives of companies when choosing an issue method. United States based companies almost solely issue equity which is full underwritten, whereas in the United Kingdom rights issues are more common (Burton & Power, 2003). The negative reaction to equity issues is, at first, surprising since according to basic corporate Finance theory companies only want to raise money when they have access to positive NPV projects (Armitage, 1998). Leading explanation for the negative reaction is the model by Myers & Majluf (1984), which explains the negative reaction by stating managers know more about the company than investors, and will subsequently issue shares when they are overvalued (Armitage, 1998). Statistical evidence in the paper of Armitage is contradictive compared to Burton, Lonie & Power’s (1999) paper, stating that rights issues have a less negative effect than other seasoned equity offerings (Eckbo & Masulis, 1992)(Armitage, 1998).
As mentioned above, the Myers and Majluf model (1984) for explaining negative abnormal returns around the announcement of a seasoned equity offering could be caused by the fact that managers know more about the company than the market does. When they ‘time’ their issues right (only issue when overvalued), they can capture more money compared to not timing their issue. Assuming that share prices do not reflect the true value of the company but are unbiased, half of the companies will have overvalued stock and the other half will have undervalued stock. But when companies only issue equity (or, at least, most of them) when shares are overvalued, the market will desire a compromise in the form of a discount, explaining the negative reaction the market has to seasoned equity offerings (Armitage, 1998).
7
Purnanandam & Kim (2006) empirically tests theories about the negative market reaction to seasoned equity offers. Understanding what theory could best explain the market’s reaction is crucial when drawing conclusions from regressions, as well as uncovering variables which could be added to the model, adding more explanatory power. Purnanandam & Kim (2006) investigate three theories presented by other researchers, namely Leland and Pyle’s (1977) signaling effect theory, Myers & Majluf’s (1984) adverse selection problem theory and Jung, Kim and Stulz (1996) agency problems theory. Leland & Pyle (1977) attribute the negative reaction to seasoned equity offerings to the sale of shares by better-informed investors, which signals shares are overvalued (Purnanandam & Kim, 2006). Myers & Majluf’s (1984) theory extents this theory by suggesting that ‘the mere act of issuing equity conveys a negative signal about the true value of the firm’ (Purnanandam & Kim, 2006). Jung, Kim and Stulz’ (1996) theory attributes the negative market response to misalignment between managerial self-interest and shareholder value maximization (Purnanandam & Kim, 2006). In Purnanandam & Kim’s (2006) paper, empirical confirmation for both Leland & Pyle’s (1977) theory and Jung, Kim & Stulz theory is found. Issuance of non-rights offerings does not create negative abnormal returns according to Purnanandam & Kim’s (2006) research, which led Purnanandam & Kim into questioning Myers & Majluf’s (1984) theory. Not the mere act of issuing equity causes the share price to drop, but issue method seems to have influence on the share price.
Corwin (2003) investigated the loss in proceeds due to underpricing of seasoned equity offers. Seasoned equity offerings are underpriced due to the fact that information asymmetry and price pressure play a role in the market’s reaction to seasoned equity offerings (Corwin, 2003). Akerlof (1970) explains in his ‘market for lemons’ model why cars are underpriced. The rational explanation for underpricing lies in the information
asymmetry between buyers and sellers (Akerlof, 1970). This explanation can partly explain the negative reaction to seasoned equity offerings. When firm value uncertainty gets higher, the negative reaction to a seasoned equity offer gets bigger, which is in line with Akerlof‘s (1970) theory (Corwin, 2003). Underpricing is of great importance for managers and
investors, since the subsequent average loss in proceeds due to underpricing is a significant part of total direct and indirect costs of issuing, amounting to a total of 21.7% of the total cost of issuing in 1998 (Corwin, 2003).
8
Cornett et al. (2014) have researched the effect of abnormal returns of seasoned equity offerings during the crisis. Cornett et al. (2014) find a significant negative abnormal return during the crisis over a three-day period around the announcement date of a seasoned equity offering, as opposed to a non-significant abnormal return for pre-crisis equity issues. For this research, they investigated 129 seasoned equity offerings issued by financial companies and 356 seasoned equity offerings by non-financial companies during the period from 2002 to 2013. A separation between the categories financial and non-financial
companies is made because the purpose of the equity issuance may differ between firms. For example, some companies might issue equity due to capital requirements as opposed to funding needs (Cornett et al., 2014).
3 Data
Before the methodology and results are discussed, a small summary of the research will be presented stating which countries are used ,what methodology is used to test the research question and which datasets are used to complete the research.
The delineation of my dataset is based on existing literature, on regulation on the issuing process and partly on data-availability. Eckbo & Masulis (1992) state that rights offers have nearly disappeared in the United States. Looking in the Thomson-one database reveals this statement still holds, with only four available announcements available regarding rights issues during the crisis. For this reason, United States is not included in the dataset. Looking in the Thomson-one database also reveals that both rights and non-rights offerings are common in Western-Europe. As discussed in the literature review, the United Kingdom has different rules concerning seasoned equity offerings. Shareholders in the United Kingdom have a pre-emptive right to shares (Armitage, 1999). The newest act in UK’s law is the one from 2006, which also includes shareholders’ pre-emptive right. In recent years, almost all data available regarding seasoned equity offering announcements are rights issues, making it hard to compare to non-rights issues. If an effect is found on rights issues, it might be because of the fact that issues in the United Kingdom have a different effect on abnormal returns as opposed to the other countries in Western Europe. For these reasons, the United
9
Kingdom is not part of the dataset. The definitive region that has been investigated is Western-Europe, excluding the United Kingdom. The time-span for which seasoned equity offerings are investigated will be from 2008 until 2015. According to literature by Cornett et al.(2014), a negative abnormal return should be expected for seasoned equity offerings during the crisis.
In this research, a market model is used which makes use of cumulative abnormal returns. For the calculation of abnormal returns an event window and estimation window has to be set up. The event date is evidently the announcement date of a seasoned equity offering. The abnormal returns are calculated in the event window, two days before the
announcement date until two days after the announcement. The two day period prior the announcement date is added to capture any rumors regarding the issue of equity. Adding more days before or after the announcement might capture the abnormal returns for other events and is therefore not included when estimating the abnormal returns.
To complete the empirical research, three datasets are necessary. The first dataset contains information about seasoned equity offer announcement dates per company. The second dataset is needed to look-up share prices for said companies. Finally, for the calculation of the model, a benchmark representative for European companies is needed . By combining these three datasets, I should ultimately find the abnormal returns.
The first database I use is the Thomson-one database (formerly known as SDC platinum). This database provides data about announcement dates of seasoned equity offerings. The Thomson-one database also provides company details such as macro-industry of the company, which will be used to find differences between industries. The second database used is the Datastream database, which will provide stock prices on which all market parameters will be based. Next to it, Datastream also provided the benchmark used to calculate the market parameters. The dataset used as a benchmark is the EUROSTOXX 50, which contains fifty blue-chip companies based in Europe.
The companies used are all Western-Europe based (excluding the United Kingdom), which is the only criteria used to look up companies in this time period. A potential exclusion bias exists in the fact that announcement dates are not readily available for every SEO. Only for 1% of seasoned equity offerings, an announcement date is available. Some countries might be overly represented due to the fact that these countries have a higher announcement to
10
SEO ratio than others, which may causes biases towards results that are more
representative for these countries than Europe as a whole. An initial sample of 86.149 equity issues between August 1st 2008 and June 1st 2015 is obtained using the Thomson-one database. Follow-ons in Europe including an announcement date narrows the scope of data down to only 3810 dates. Narrowing this sample down to Western-Europe excluding the United Kingdom brings the total to 277. Two pools of data are to be extracted from this sample, namely rights and non-rights offerings. In total, the two datasets contain 120 rights offerings and 421 non-rights offerings are the subsequent pools of data. These data pools contain some form of ‘noise’. This noise is mostly the non-availability of stock prices of some companies. After correcting for noise, I obtain 36 rights offerings. For the non-rights dataset, the 50 biggest companies with data available where chosen. After calculating all market parameters, one company was left out of this group, totaling the number for the non-rights dataset to 49. For each of these companies, the stock prices are obtained as well as the industry the companies are active in. These 85 companies can also be divided over macro industry. For this research, the companies are divided into four groups of industry, namely financial, high technology, industrial and others. The groups contain 21, 13, 20 and 31 companies respectively.
3.1 empirical methodology
In the next two subsections the model estimated in this thesis will be presented followed by the methodology. To test the research question stated at the beginning of this thesis, two hypotheses are formulated. The model estimated in this thesis is first presented for a better understanding of the hypotheses.The model estimated to test the hypotheses is the
following:
In this model, is a dummy variable for rights issues, where equals one when the company issued equity in the form of a rights offering. , and
11
equal one when a firm is in the financial, industrial or in the high technology industry respectively. If a firm operates in a different sector than these three, the dummy variables ,
and will all equal zero.
3.2 methodology
This part contains the papers some assumptions are based on and the model used to estimate the model presented in equation (1). The model used to estimate the model presented in equation (1) is the market model, which can be used to perform an event study. The market model assumes a linear relationship between market return and performance of the company examined. The first thing to do when conducting an event study is defining the event and identifying the period over which the firms will be examined (Mackinlay, 1997). The period over which the firms were examined is the first of August 2008 until the first of June 2015. In this time-window, stock prices of companies who have made a seasoned equity offer are examined by taking the stock prices from six months prior the announcement date until one month prior the announcement date. The benchmark which the betas are calculated against is the EUROSTOXX 50, which consists of fifty blue-chip Europe based companies. Based on this information the beta of a company is calculated, which is necessary for the calculation of abnormal returns. In this research, a market model is used which makes use of cumulative abnormal returns. For the calculation of abnormal returns an event window and estimation window has to be set up. The event date is evidently the announcement date of a seasoned equity offering. The abnormal returns are calculated in the event window from two days prior the announcement date until two days after the announcement. The two day period prior the announcement date is added to capture any rumors regarding the issue of equity. Adding more days before or after the announcement might capture the abnormal returns for other events and is therefore not included when estimating the abnormal returns.
3.3 hypothesis
12
econometrically testable, which is necessary to answer the research question. The second set of hypotheses are formulated in a single sentence for convenience of the reader, to better understand what the two hypotheses state.
The hypothesis are:
For convenience, the two hypothesis are also stated in a single sentence form: : The abnormal returns are not dependent on the equity issue method : The abnormal returns are dependent on the equity issue method
3.4 Tests and model
In this subsection, the papers and methodology used to estimate the model are described. The model estimated to test the hypotheses is presented as equation (1) in this thesis. Next up the market model will be discussed, which is used to estimate the average abnormal returns and cumulative average abnormal returns. The assumptions and methodology regarding the market model are broadly discussed.
The abnormal return is the difference between the expected market response and the actual market response, which is measured using the formula
( | )
where is the abnormal return at time τ , is the actual return at time τ and
( | ) is the normal return at time τ . The market model assumes a linear relationship
between market return and the return of a security (Mackinlay, 1997). The abnormal return calculates the difference between expected return, calculated against a benchmark, and the actual return. The reaction of a firm measured against the market is called the beta,
13
which is used to calculate said expected return. The beta will be calculated against a benchmark over the estimation window, where the beta will be calculated between six months before the announcement date and one month before the announcement date. After calculating the beta, the market model for any security i is
( ) ( )
where and are the returns for company i and the market portfolio in period t
respectively, and is the disturbance term with zero mean. The market parameters are
given by the symbols , and where is the alpha of company i, is the beta of company i and is the variance of the error term of company i (mackinlay, 1997). The benchmark used to represent the market portfolio return is the EUROSTOXX 50, which contains fifty blue-chip companies in Europe.
For the measurement of abnormal returns, some notations need to be made to clarify some symbols. Returns will be indexed in event time using τ . Some quantitative measures are necessary to facilitate the formulas used in this thesis. First off all, the event date is defined as τ=0. The event window is defined as , and the estimation window is defined as (Mackinlay,1997). The length of the estimation window and event window can then be specified as and respectively. The event window in this thesis is two days prior the announcement date and two days after the announcement date.
Assumptions made for this model are that all variables are independently and identically distributed random variables and also multivariate normally distributed. Assuming these assumptions hold, OLS is efficient (Mackinlay, 1997). Under general conditions OLS is also a consistent estimator (Mackinlay,1997). The calculation of the market parameters are
described in the appendix, under appendix 1. The statistical properties of the OLS estimators are explained in the following paragraph. After calculating the market model parameters, the abnormal return can be calculated using (2) (Mackinlay,1997).
14
When is the sample of for firm i in the event window, the
abnormal return can be calculated using the sample formula , which is (Mackinlay, 1997):
̂ ̂
The abnormal return measures the difference between the expected return, measured by the estimated market parameters described in the appendix, and the actual return. Under the null hypothesis, conditional on the event window market returns, the abnormal returns will be jointly normally distributed with a zero conditional mean and conditional variance
( ) where
( ) [
̂
̂ ]
is the conditional variance (Mackinlay, 1997). The conditional variance consists of two parts. The first part of the conditional variance is , which is indicated in (3), measures the
variance of the errors. The second part is added due to sampling errors in and . This sampling error leads to serial correlation of the abnormal returns (Mackinlay, 1997). As already presented, is the event window. When said event window becomes large, the second part of the conditional variance goes to zero since the errors made by measuring the market parameters become smaller. The variance of the abnormal return will be and the abnormal return observations will become normal over time (Mackinlay, 1997). For this reason, an event window has to be chosen such that the second part of (7) becomes close to zero. If L1 is too small, and adjustment has to be made regarding the standard errors of
abnormal returns. According to Mackinlay, it is reasonable to assume that the second part becomes zero when is large enough (Mackinlay, 1997).
Under the null hypothesis, the distribution of the sample abnormal returns of a given observation in the event window is (Mackinlay, 1997):
15
Equation (6) is used for the aggregation of abnormal returns. ( ) is used as the
sample variance of stock i’s abnormal return, . According to Mackinlay (1997),
aggregation goes across the dimensions time and securities. First, aggregation across time is discussed followed by aggregation across securities. The aggregation of abnormal returns is often referred to as cumulative abnormal returns, or CAR. Let CARi( 1, 2) be the sample
cumulative abnormal return from 1 to 2 , where T1< 1≤ 2≤T2. The cumulative abnormal
return is then the sum of all included abnormal returns of a given firm in time-period , formulated by:
∑
The variance of is given by the equation
This sample estimate can be used for reasonably large samples of L1. If L1 is too small, an
adjustment in the variance of CAR has to be made for larger errors in the market parameters. For this reason, and for the reason provided in (5), a large number for L1 is
taken for the calculation of the market parameters. L1 is, as said in the introduction of the
research plan, 110 days for all market parameter calculations (companies’ betas are calculated over these 110 days minus holidays where the stock exchange is closed). Under the null hypothesis, the distribution of CAR is
( )
Given these equations, test on abnormal returns can be done. Aggregation across time has been discussed, now aggregation across securities will be of order. For the model to be
16
independent across securities the event window of different securities can’t overlap. The phenomenon of overlapping event windows is called clustering (Mackinlay, 1997). Using equation (6), the individual abnormal returns can be aggregated (Mackinlay, 1997). To calculate the average abnormal return across securities, the following equation is derived from (6)
̅̅̅̅ ∑
,where N is the number of events and ̅̅̅̅ is the average abnormal return for a security in time-period . The variance for the average abnormal returns is, if L1 is large enough, given
by the equation
̅̅̅̅ ∑
The average abnormal returns can now be calculated and tested. Two final equations have to be explained, since tests on cumulative average abnormal returns are also discussed in this thesis. The equation for the cumulative average abnormal return, or ̅̅̅̅̅̅, is ̅̅̅̅̅̅ ∑ ̅̅̅̅
The ̅̅̅̅̅̅ follows the following distribution:
( ̅̅̅̅̅̅ ) ∑
Based on these equations, the average abnormal returns and cumulative average
abnormal return can be calculated. The results are presented in section 4 of this thesis, and conclusions about the results are given in section 5 of this thesis.
17 4 Results
Using the market model explained in section 3.4, the market parameters and subsequently the abnormal returns can be calculated. To prevent the event itself from having an effect on the market parameters, the month prior to the announcement date is excluded for the calculation of betas. After calculating all abnormal returns, the model presented in section 3.1 can be estimated and measured for significance. For all companies the output on the beta and alpha calculation looks like this:
In this example, the beta is 1.14 and the alpha is -0.001. The mean squared residual is used to calculate the standard deviation of the cumulative average abnormal return, using the Mackinlay method as described in section 3.4. The number of observations in this example is 109. The estimation window is from six months prior the announcement date until one month prior the announcement date. Weekend days and (some) holidays are not
incorporated, since the stock exchange is closed in the weekends and on (some) holidays. This structure is used to calculate all alphas and betas.
The market parameters are then used to calculate the abnormal returns per firm. Below is an example on how the abnormal returns are calculated based on equation (2). For all companies, abnormal returns are calculated in this manner.
18
The columns (1) and (2) are the prices of the company and market respectively, while columns (3) and (4) display the returns for the company and the market respectively. Under column (5), the abnormal return is calculated using the alpha, beta and returns calculated under column (3) and (4). After calculating the market parameters and subsequently the abnormal returns, the results can be tested. The results for rights-offerings are as follows:
Under ̅̅̅̅ , the average abnormal returns per day are given. AR -2 stands for two days before the announcement is made, AR -1 for one day before the announcement is made etcetera. ̅̅̅̅̅̅ stands for cumulative average abnormal return, which is the mean of all accumulated abnormal returns per company. Under std. Deviation, the standard deviation
19
of the average abnormal returns per day are given. The numbers under t are significance tests, calculated using a one-sample t-test. The t-test is
̅
√ ⁄
where ̅ is the average abnormal return, is value of the average abnormal returns under the null hypothesis, S stands for the standard deviation and n stands for the number of observations. The numbers of t for significance on a 10%, 5% and 1% level are (-)1.671, (-)2 and (-)2.66 respectively. Before comparing the two groups, they are first looked at
individually, beginning with the rights offerings.
Looking at the individual days, all days except two days after the event date, have an abnormal return significantly different from zero on a 1% level. This means that, based on the sample, proof is found for an abnormal return significantly different from zero for all days which are investigated except the day two days after the event date.
Looking at ̅̅̅̅̅̅, the average accumulated abnormal return is -0.8%. This result is, after testing with the t-test, significantly different from zero on a 1% significance level using the Mackinlay standard errors. This means that, based on the sample, rights offerings in Europe during the crisis have a significant negative effect on share prices accumulated over a five-day period.
20
In the non-rights sample, looking at the days individually, all days have a return significantly different from zero on a 1% significance level except the day two days before the event date. What immediately stands out is the difference in abnormal returns for non-rights offerings compared to rights offerings around the event date. For rights issues, the one-day period surrounding the event date all abnormal returns are negative, as opposed to non-rights offerings, where for the one-day period surrounding the event date all abnormal returns are positive. The cumulative average abnormal return for the non-rights sample is 0.3%, and is significant at a 1% significance level.
Another test is conducted using Stata, to check whether or not industry is an explanatory variable for abnormal returns. Regressing the cumulative average abnormal return on the variables , and does not give a significant result for explaining the cumulative average abnormal return. Regressing the event date on the industry variables does give a significant result for the variable , with a p-value of 0.071 and a beta of 0.02 as can be seen in table 5
21
When a regression is done using the event date abnormal returns as the y variable and the variable as the x variable, the outcome is even more significant with a p-value of 0.056 (see appendix 2).
The final two regressions are regressions where the dummy is used as an
explanatory variable for abnormal returns. The results differ from the results presented in the beginning of section 4, since regression uses a different standard deviation compared to the ones used earlier in this thesis, which are based on Mackinlay’s market model. The main difference is that Mackinlay’s method uses standard deviations based on the mean square residual when beta and alpha were calculated, and Stata gets its standard deviation from the list of cumulative abnormal returns. The output of the regression in Stata is as follows:
22
As can be seen in table 6, the p-value is 0.6, which indicates an insignificant effect of the dummy on ̅̅̅̅̅̅. This result is contradictive with results presented earlier. As stated before, the difference occurs due to a difference in standard deviation calculation. The last
regression used for this thesis is the regression of announcement date abnormal return plus the day after announcement date abnormal return on the rights variable. The reason for this regression is that these two days tend to have the biggest effect on stock prices, and
subsequently, the biggest difference between both methods should be found. The regression gives the following output:
23
This regression gives a p-value of 0.15 which is, although better than the previous regression, still insignificant. Again, this insignificant result comes from the different standard deviation regression uses as opposed to the Mackinlay method.
5 Conclusion
The research question stated at the beginning of this thesis is ‘’do rights issues have a different market reaction than non-rights issues during the crisis?’’. Looking at the regression table 6, the answer to this question is no. Looking at the t-statistics calculated using the standard deviations as seen in tables 3 and 4 , the answer to this question is yes. The reason these two results are contradictive lies in the different standard errors used. Using the regression, the H0 hypothesis is not rejected. Using the tables with Mackinlay
standard errors, the H0 hypothesis is rejected. Tests done on industry dummy variables show
no significant effect of industry on explaining abnormal returns for the financial industry and the high technology industry. However, for the industrial firms, the dummy variable is significant on a 10% level with a p-value of 0.071. Comparing the results with existing literature reveals that this research corresponds best with the paper written by Burton, Lonie & Power(1999), stating rights offerings have a more negative effect than non-rights offerings. However, Burton, Lonie & Power (1999) state a negative effect for both groups, as well as far more negative results compared to the results found in this thesis. The difference may come from the sample used to answer the research question. In Burton, Lonie &
Power’s (1999) research, only the United Kingdom was investigated, which has different laws regarding equity offerings compared to the rest of Europe. Another difference between this thesis and the paper written by Burton, Lonie & Power (1999) is the fact that all
offerings for the sample of this research are from the period 2008 to 2015, which is a period of financial crisis contrary to the period researched by Burton, lonie & Power (1999). For further research it would be interesting to investigate differences between the United Kingdom and the rest of Western Europe within the same time-period, and check if the large negative abnormal returns found by Burton, Lonie & Power (1999) for the UK market still hold.
24
6 References
Akerlof, G. A. (1970). The Market for "Lemons": Quality Uncertainty and the Market Mechanism. The
Quarterly Journal of Economics, 84(3), 488-500.
Armitage, S. (1998). Seasoned equity offers and rights issues: a review of the evidence. The European
Journal of Finance, 4(1), 29-59.
Armitage, S. (2012). The calculation of returns during seasoned equity offers. The European Journal
of Finance, 18(5), 393-417.
Asquith, P., & Mullins, D. W. (1986). Equity issues and offering dilution. The journal of Financial
Economics, 15(1-2), 61-89.
Bhandari, L. C. (1988). Debt/Equity Ratio and Expected Common Stock Returns: Empirical Evidence.
The journal of Finance, 43(2), 507-528.
Burton, B. M., Lonie, A., & Power, D. M. (1999). Does the issue method influence themarket reaction to seasoned equity offerannouncements? Applied Economics Letters, 6(7), 459-462.
Burton, B. M., & Power, D. M. (2003). Evidence on the determinants of equity issue methodin the UK. Applied Financial economics, 13(2), 145-157.
Bøhren, Eckbo, B. E., & Michalsen, D. (1997). Why underwrite rights offerings? Some new evidence.
The journal of Financial Economics, 46(2), 223-261.
Cornett, M. M., Mehran, H., Pan, K., Phan, M., & Wei, C. (2014). CDS and Equity Market Reactions to Stock Issuances in the U.S. Financial Industry: Evidence from the 2002-13 Period. Federal
Reserve Bank of New York Staff Reports, 697, 1-43.
Corwin, S. A. (2003). The Determinants of Underpricing for Seasoned Equity Offers. The Journal of
Finance, 58(5), 2249-2279.
Eckbo, B. E., & Masulis, R. W. (1992). Adverse selection and the rights offer paradox. Journal of
Financial Economics, 32(3), 293-332.
Jung, K., Kim, Y. C., & Stulz, R. M. (1996). Timing, investment opportunities, managerial discretion, and the security issue decision. The Journal of Financial Economics, 42, 159-185.
Kraus, A., & Litzenberger, R. H. (1973). A State-preference Model of Optimal Financial Leverage. The
journal of Finance, 28(4), 911-922.
Leland, H. E., & Pyle, D. H. (1977). Informational Asymmetries, Financial Structure, and Financial Intermediation. The Journal of Finance, 32(2), 371-387.
Mackinlay, A. C. (1997). Event Studies in Economics and Finance. Journal of Economic Literature,
25
Marsh, P. (1982). The Choice Between Equity and Debt: An Empirical Study. The journal of Finance,
37(1), 121-144.
Masulis, R. W., & Korwar, A. N. (1986). Seasoned equity offerings an empirical investigation. Journal
of Financial Economics, 15(1-2), 91-118.
Modigliani, F., & Miller, M. H. (1958). The Cost of Capital, Corporation Finance and the Theory of Investment. The American Economic Review, 48(3), 261-297.
Myers, S. C., & Majluf, N. S. (1984). Corporate financing and investment decisions when firms have information that investors do not have. The journal of Financial Economics, 13(2), 187-221. Ngatuni, P., Capstaff, J., & Marshall, A. (2007). Long-Term Performance Following RightsIssues and
Open Offers in the UK. Journal of Business Finance & Accounting, 34(1), 33-64.
Purnanandam, A. K., & Kim, E. H. (2006). Why Do Investors React Negatively to Seasoned Equity Offerings? Ross School of Business Paper, 1053, 1-53.
Schipper, K., & Smith, A. (1986). A comparison of equity carve-outs and seasoned equity offerings: Share price effects and corporate restructuring. The journal of Financial Economics, 15(1-2), 153-186.
26
7 Appendix
Appendix 1
For the firm in event time, the OLS estimators of the market model parameters for an estimation window of observations are
̂ ∑ ̂ ̂ ∑ ̂ ̂ ̂ ̂ ̂ (4) ̂ ∑ ̂ ̂ (5) where ̂ ∑ and ̂ ∑
As stated under (2) the companies return and the market return at time t are and respectively. The variables ̂ and ̂ are the averages in the estimation period for the
27
Appendix 2
Announcement date on industrial dummy regression.
