The announcement effect on mean and varance for underwrtten and non-underwrtten SEOs Bachelor Essay n Fnancal Economcs Department of Economcs Sprng 013 Marcus Wkner and Joel Anehem Ulvenäs Supervsor: Professor Ola Bengtsson 013-05-30
Abstract Ths thess nvestgates the stock return and ts varance around seasoned equty offerng announcements for Swedsh companes lsted on the OMX Large cap, Md cap and Small cap exchanges. The analyss s made on a full sample contanng 5 SEOs, as well as two subsamples contanng underwrtten and non-underwrtten SEOs. The framework for the event study s OLS regressons based on the CAPM-model. Durng the studed sample perod, January 006 to December 010, companes makng SEO announcements are found to exhbt a sgnfcant negatve average cumulatve abnormal return of around.5 percent on the announcement day as well as for a three-day horzon. For longer horzons, the average cumulatve abnormal return s around negatve 1.6 percent. Non-underwrtten SEOs are found to exhbt less negatve returns than underwrtten ones, whch s n lne wth prevous studes nvestgatng ths matter. The return varance for an ssung company s found to ncrease durng the month followng the SEO announcement for 40 out of 5 companes, whereof 30 varance ratos are found to be sgnfcant. Further, there s no evdence that there s a sgnfcant dfference n return varance between underwrtten and non-underwrtten SEOs. 1
Table of Contents Abstract... 1 Word lst... 4 1.1 Introducton... 5 1. Purpose... 5 1.3 Problem dscusson... 6 1.4 Delmtatons... 7 1.5 Outlne... 7. Prevous research, SEO ssung process & benefts of underwrtng.... 8.1 Prevous research... 8. The SEO ssung process... 9.3 The Benefts of underwrtng... 9 3. Theory... 11 3.1 The peckng order theory for rasng captal... 11 3. Theoretcal dscusson about asset prcng models... 13 3.3 Expected answers to stated hypotheses based on economc theory... 15 4. Data and Varables... 16 4.1 Data sample... 16 4. Delmtatons n the data sample... 17 4.3 Varables... 17 4.3.1 The rsk free nterest rate... 17 4.3. Return varables... 17 4.4 Testng the varables for statonarty... 18 5. Method... 19 5.1. Event study methodology... 19 5. Estmaton Wndow... 0 5.3 Event Wndow... 0 5.4 Measurng normal returns... 0 5.5 Measurng abnormal returns... 1 5.6 Dagnostc checkng of the regresson model... 1 5.7 Testng for Heteroskedastcty... 5.8 Testng for Autocorrelaton...
5.9 Jarque-Bera normalty test... 3 5.10 Correctng for heteroskedastcty and autocorrelaton... 4 5.11 Testng the sgnfcance of average CAR... 4 5.1 Testng for equalty of sample means... 5 5.13 Varance tests... 6 5.14 Relablty... 7 5.15 Valdty... 7 6. Results of the event study... 9 6.1 Results regardng returns, full sample... 9 6. Results regardng returns, sub sample... 30 6.3 Results regardng varance, full sample... 30 6.4 Results regardng varance, sub samples... 31 7. Analyss and dscusson... 3 7.1 Answers to hypotheses 1-4.... 3 7. Results n lne wth theory?... 3 7.3 Concludng dscusson... 3 8. References... 34 9. Appendx... 37 3
Word lst Seasoned Equty Offerng (SEO) = Nyemsson Underwrtten SEO = Garanterad nyemsson Non-underwrtten SEO = Ej garanterad nyemsson Underwrter = Garant Drected share ssue = Rktad nyemsson Rghts ssue = Nyemsson med företrädesrätt Subscrpton rght = Tecknngsrätt Subscrpton prce = Tecknngskurs Prmary preferental rght = Prmär företrädesrätt Subsdary preferental rght = Subsdär företrädesrätt Dluton = Utspädnng Prospectus = Emssonsprospekt Record date - Avstämnngsdag Flotaton cost = Total emssonskostnad Take-up level = Tecknngsgrad Cumulatve abnormal return (CAR) = Kumulatv överavkastnng 4
1.1 Introducton Seasoned equty offerngs have been frequently occurrng on the Swedsh stock market, especally durng the fnancal crss of 008-010. Underwrtten SEOs has gone from beng less frequent to representng a majorty of the performed SEOs. The hgh provson pad to underwrters for guaranteeng full subscrpton of the SEOs has been wdely debated n fnancal meda and questoned by many. 1 The flotaton cost assocated wth SEOs can be weghed aganst the nterest expense a company would pay f t chose bank loan fnancng, or the nterest rate t would pay on a corporate bond f t would chose a bond ssue. Just as the nterest expense, the flotaton cost mpacts the corporate value negatvely. Accordng to Eckbo and Masuls (199), underwrter compensaton accounts for approxmately 90 percent of total flotaton costs, makng them by far the most sgnfcant cost n the ssung process. Mnmzng flotaton costs should therefore be of nterest to all exstng shareholders, and t logcally follows that mnmzng underwrter compensaton should be the man target. 1. Purpose The purpose of ths thess s to nvestgate whether underwrtten SEOs have sgnfcantly better return propertes than non-underwrtten SEOs. In ths thess, better s defned as hgher abnormal returns coupled wth lower varance, measured over the event wndow. Accordng to the purpose, the followng hypotheses are constructed: Hypothess 1: Is there sgnfcant abnormal return around SEO announcements? Hypothess : Is there a sgnfcant dfference n cumulatve abnormal returns between underwrtten SEOs and non-underwrtten SEOs? Hypothess 3: Is there, n general, an ncrease n return varance for the ssung company durng the month followng the SEO announcement? Hypothess 4: Do underwrtten SEOs exhbt lower event wndow varance than non-underwrtten SEOs? 1 For example, by former presdent of Aktespararna, Günther Mårder, 010. 5
1.3 Problem dscusson Wthout the use of underwrters, a company ssung new shares as a tool to rase captal faces the rsk of not recevng the requested funds. Ths shortfall n captal occurs f exstng shareholders do not use all ther subscrpton rghts to buy the newly ssued shares. If a company experences a shortfall of captal, the board of the company usually has a strategy for gong forward. The man solutons usually undertaken by managers are to extend the rghts ssue perod, seek alternatve means of fnancng, or proceed wth current operatons at a slower pace. To address the problem of possble captal shortfall, companes use underwrters n order to ensure that ther SEO get fully subscrbed. The provson pad to underwrters s typcally calculated as a percentage of the underwrter s guaranteed amount. The percentage rate usually falls wthn the range of 0-10 percent, wth hgher percentages for companes wth smaller market captalzaton and smaller percentages for companes wth larger market captalzaton and fnancal nsttutons, e.g. banks. Accordng to the above, the underwrtng process can be regarded as an nsurance polcy. However, t s not perfectly clear whether the underwrters oblgaton to actually complete the subscrpton s legally bndng. 3 There have been cases when the underwrters have not fulflled ther oblgatons due to personal bankruptcy or bankruptcy of the underwrtng company. It s thus ultmately the responsblty of the ssung company to ensure that the credt worthness of underwrters s suffcent to fulfll the stated oblgatons. For the ssung company, payng the underwrter can be regarded as buyng a put opton on ts own stock, snce the ssung company buys the rght to sell stock at a pre-determned prce. 4 A long put opton n combnaton wth exstng stock s called a protectve put and have the beneft of reducng rsk for ts holder snce ts payoff structure reduces the varance of cash flows, especally when the strke prce s below the exercse prce. 5 It s therefore of nterest Gustavsson, M and P. Lndström, 010, Garanter vd nyemssoner förutsättnngar och kostnader. 3 Hoffman Bermejo and Raudsepp (009) states that underwrtng agreements are not legally bndng. Professor n Swedsh and nternatonal busness law, Erk Nerep, s of the opnon that underwrtng agreements are legally bndng. Prawtz (009) dscusses jurdcal arguments both for and aganst the legally bndng ssue and calls for legslatve authortes to provde a clear statement that settles ths queston once and for all. 4 Many studes valung the underwrter agreement usng optons theory has been made. See for example research made by Marsh, 1994, Underwrtng of Rghts Issues, a study of the returns earned by sub-underwrters from UK rghts ssues and Marsh, 1998, Sub-underwrtng of rghts ssues, a falure of competton? 5 Bode, Z., A. Kane and A.J. Marcus, 005, Investments, Sxth Edton, McGraw-Hll, Sngapore, p. 711. 6
to study whether or not underwrtten SEOs have the beneft of lowerng the event wndow varance, compared to non-underwrtten SEOs. 1.4 Delmtatons Subscrpton commtments made by large shareholders has not been regarded as underwrtng. Only the percentage of the SEO that has been guaranteed by underwrters earnng provson has been counted as a guarantee. The study s lmted to rghts ssues wth prmary preferental rghts. No analyss regardng the underwrters jurdcal responsbltes s undertaken. It s noted that uncertanty regardng underwrter agreements jurdcal mplcatons s prevalng, but no attempt of resolvng ths problem has been made. The event study s based on the CAPM-model only. Only Swedsh market data has been used for regresson estmatons. No modelng of what affects the sze of the underwrter fee s ncluded. 1.5 Outlne The paper s organzed as follows. Secton presents the results from prevous research, gves an ntroducton to the SEO ssung process and descrbes the benefts of underwrtng. Secton 3 outlnes the theoretcal framework. Secton 4 descrbes the data and the varables used n the regressons. Secton 5 descrbes the event study methodology and the test procedures as well as dagnostc checkng of the estmatons. Secton 6 contans the results of the study and secton 7 concludes. 7
. Prevous research, SEO ssung process & benefts of underwrtng..1 Prevous research Many studes regardng announcement effects around SEOs have been conducted for the Swedsh stock market. Malmström and Nlsson (00) study the sample perod 1993-001, and fnd postve cumulatve abnormal returns. 6 They also examne whch frm specfc varables are sgnfcant determnants of the abnormal returns. Von Arronet, Källstrand and Tarnawsk-Berln (003) compare announcement effects between sectors n Nordc countres. 7 Frtzell and Hansveden (006) study the sample perod 1986-005, and fnd a negatve CAR of around percent on the event day. 8 Egerot, Hagman and Svensson (009) use the sample perod 1997-008 and fnd a negatve CAR for all ther examned event wndow horzons. 9 Gustavsson and Lndström (010) nvestgate whch factors cause the decson of usng underwrters. 10 Månsson and Rostedt (010) analyse the effect of SEO announcements on returns dependng on the purpose of the SEO, debt payback, acquston, or ncrease of workng captal, and fnd negatve CARs for all purposes. 11 To our knowledge, only one study wth focus on comparng return propertes for underwrtten SEOs and non-underwrtten SEOs has been made. Andersson and Söderberg (007) studes the sample perod 1986-005, wth focus on abnormal returns and the offerng dscount. The authors fnd that SEOs n general exhbts a negatve average CAR of around percent on the announcement day and a negatve 1 percent for other horzons. 1 They also fnd that non-underwrtten SEOs exhbt less negatve average and medan CAR than underwrtten SEOs for all event wndow horzons. The authors also perform a crosssectonal analyss of the abnormal returns n order to determne whch frm specfc varables are sgnfcant factors for explanng CARs. The Andersson and Söderberg 6 Malmström, K. and A. Nlsson, 00, Annonserngseffekt av nyemssoner - En fallstude på Stockholmsbörsen. 7 Von Arronet, C., J. Källstrand, and M. Tarnawsk-Berln, 003, Kursreaktoner på tllkännagvande av nyemsson. 8 Frtzell, M. and J. Hansveden, 006, Stock Market Reactons and Offerng Dscounts of Swedsh Equty Issues. 9 Egerot, R., E Hagman, and M. Svensson, 009, Deltagande I nyemsson - en buy and hold-strateg. 10 Gustavsson, M and P. Lndström, 010, Garanter vd nyemssoner förutsättnngar och kostnader. 11 Månsson, M. and C. Rostedt, 010, Varnng för ras En stude av aktemarknadens reakton på nyemssonsbeskedet. 1 Andersson, M.E. and S. Söderberg, 007, Rghts Issues n the Swedsh Market, A Comparson between Insured and Unnsured Rghts Issues. 8
research s very nterestng snce t s a predecessor to our analyss. However, the authors do not focus on a comparson regardng the varance. Our analyss, usng the sample perod 006-010, can be seen as a complementary study, wth the addtonal feature of varance comparson.. The SEO ssung process Snce the ssung of new stock s a rather complcated and tme consumng task for a company to undertake, a short ntroducton of the ssung process s ntally presented. Frst, there are dfferent ways a company can formulate the share ssue. It can choose to perform a drected share ssue, or an ssue wth prmary preferental rghts. Drected share ssues are typcally targeted to a specfc group of nvestors, often employees or nsttutons. Drected share ssues are not analyzed n ths study. In our study, only rghts ssues wth prmary preferental rghts are analyzed, snce t s the most commonly used flotaton method n Sweden. Rghts ssues wth prmary preferental rghts are drected to all exstng shareholders, who are gven rghts n proporton to ther exstng amount of stock. In case not all subscrpton rghts are used for subscrpton n the SEO, the access to the remanng subscrpton rghts s decded by the subsdary preferental rght. If there s unsubscrbed stock after both prmary and subsdary rghts have been used, the rest s subscrbed by underwrters f such has been contracted. The new shares are almost always offered at a dscount to the current market prce. The dscount s set to encourage subscrpton n the SEO. One of the most extreme examples of subscrpton prce dscounts n Sweden s the Scandnavan Arlnes, SAS, SEO n 009, where new shares were offered to the market at a 90 percent dscount. 13 In addton to the dscount, SAS used underwrters to make sure the SEO would get fully subscrbed. A large dscount puts more value n the subscrpton rght and causes a larger dluton of the stock prce. Underwrters are subscrbng drectly to the ssung prce stated n the SEO prospectus. Ths s benefcal to underwrters, snce they receve a dscount n addton to ther underwrter compensaton..3 The Benefts of underwrtng The benefts of underwrtng have generally been sad to be: 13 http://www.va.se/nyheter/nyemssoner-onodgt-dyrt-53775 9
1. To gve managers a certan and stable envronment to operate n.. To show exstng shareholders, as well as the rest of the market, that the company s operatons are of economc value, worth nvestng n. 3. To support the stock prce and decrease the return varance durng the perod the SEO s performed. Reason 1 s self-explanatory. If the company s guaranteed to obtan the captal t sought for, t does not have to devote any resources for analyss and formulaton of back-up plans n the event of crss, seek alternatve means of fnancng, etc. Reason argue that the presence of underwrters should confrm that the company s future busness plans are of economc value. The reason for ths s that before an underwrter agrees to provde a guarantee, a thorough due dlgence s usually performed. If the underwrter fnds the company s plans to be unproftable, he wll most lkely not take the rsk of provdng a guarantee. The presence of underwrters can therefore strengthen confdence and encourage exstng shareholders to partcpate n the SEO. Thus, ncentves for excessve sellng should be dampened. Reason 3 puts emphass on stock prce support. One of the bggest concerns for a company ssung new shares s the problem that arses f the stock prce falls below the subscrpton prce stated n the prospectus. The most famous case llustratng ths problem s the Swedbank SEO n 009, where the stock prce fell below the subscrpton prce durng the perod shortly after the SEO announcement. When ths occurs, exstng shareholders notce that they can buy the stock cheaper n the stock exchange rather than buyng t through subscrpton n the SEO. Snce t s now unproftable to subscrbe, shareholders therefore refran from dong so. The consequence s that underwrters become forced to subscrbe to ther full share of stock at an unfavorable prce. Snce SEOs are rarely covered to 100 percent solely by underwrter agreements, a bg fall n the stock prce vastly ncreases the probablty that the SEO wll not be fully subscrbed. Although reason 1 s probably the man reason to why companes use underwrters, reason 3 s accordng to us the most nterestng to nvestgate. The argument for that s that reason 1 and are expected to ultmately show up as an effect n 3. The presence of underwrters should accordngly decrease uncertanty about the ssung company s future, whch ceters parbus, should lead to a decrease n return varance. 10
3. Theory To be able to perform our analyss, a theoretcal framework s establshed. The peckng order theory for rasng captal and ts extensons to rghts ssues s frst examned. Then, a dscusson about asset prcng models and ther mplcatons for event studes s provded. 3.1 The peckng order theory for rasng captal The peckng order theory was frst proposed by Myers and Majluf (1984). 14 Ther theory states that companes rank ther means of fnancng n the followng way: 1. Retaned earnngs. Bank loan fnancng 3. Issue of corporate bonds 4. Issue of new shares Fnancng a project wth retaned earnngs s the cheapest, smplest and thus most preferred method. Whether or not bank loan fnancng s preferred to the ssung of corporate bonds depends on a large number of parameters whch are specfc to each company. It s thus not possble to conclude than bank loans are always cheaper. Issung of new shares s the least favorable opton, snce t s assocated wth the hghest flotaton costs. Flotaton costs nclude all fees assocated wth the SEO, such as regstraton fees, advsory fees pad to nvestment banks, underwrter compensaton, etc. The peckng order theory thus states that share ssues should be avoded f the other means of fnancng are accessble. One can therefore argue that, n general, frms wthout access to better optons,.e. cheaper sources of captal, wll choose to ssue new equty. In ther 1984 paper, Myers and Majluf assume that a company faces a short-lved, but proftable, project opportunty whch requres fnancng through a share ssue. They further assume that managers have superor nformaton about the company s ntrnsc value, compared to other nvestors, and that they act n the best nterest of exstng shareholders. Managers wll therefore decde not to ssue new shares when the company s undervalued, snce ths wll only dlute the share prce further, makng exstng shareholders worse off 14 Myers, S.C and N.S. Majluf, 1984, Corporate Fnancng and Investment Decsons When Frms Have Informaton That Investors Do Not Have 11
than before the offer. Myers and Majluf argue that a company wll only ssue new shares when managers perceve the company as overvalued. Eckbo and Masuls (199) expand the peckng order theory by ncludng an analyss of share ssues under varous flotaton methods. The authors provde a theoretcal framework for underwrtten rghts ssues, nonunderwrtten rghts ssues and frm commtments. Accordng to Eckbo and Masuls, all frms optmze ther decson regardng a share ssue based on the followng decson rule: b ( c f ) 0. Where b s the net present value of the project, c s the dfference between the ntrnsc value of the shares sold to outsders and the shares market value condtonal on the ssue decson, and f s total flotaton costs. 15 Accordng to Eckbo and Masuls, only frms wth expected take-up level very close to 1 can perform a SEO wthout usng underwrters. A hgh take-up level should sgnal hgh company qualty and thereby less severe overvaluaton, snce exstng shareholders fnd t attractve to subscrbe n the SEO. In Eckbo and Masuls, non-underwrtten SEOs are also found to be assocated wth sgnfcantly lower flotaton costs. Therefore, non-underwrtten SEOs should be expected to have less negatve abnormal returns than underwrtten SEOs. Andersson and Söderberg (007) provde a reversed argument. They argue that the use of an underwrter should sgnal that the company s less overvalued. The reason s that before an underwrter decdes to provde a guarantee, a thorough due dlgence s usually performed. If the underwrter fnds the company to be overvalued, he wll probably not be wllng to provde a guarantee. Thus the presence of an underwrter are expected to serve as a certfcaton of value and sgnal hgh company qualty, meanng that abnormal returns for underwrtten SEOs should be expected to be less negatve than for non-underwrtten SEOs. 15 Eckbo, B.E. and R.W. Masuls, 199, Adverse selecton and the rghts offer paradox. 1
3. Theoretcal dscusson about asset prcng models To conduct an event study, one has to use an asset prcng model as framework for the emprcal analyss. Most asset prcng models are bult on the foundaton that nvestors should only be compensated for exposure to non-dversfable, systematc rsk. We dscuss the Captal Asset Prcng Model, CAPM, the Fama-French three-factor model, and the Arbtrage Prcng Theory, APT, before choosng the CAPM-model. The CAPM model states that the expected return of an asset should be lnearly related to ts covarance wth the market portfolo, accordng to equaton 1: Rt Rmt t (Eq.1) Where Rt s the excess return of a certan asset, at tmet, s the regresson ntercept coeffcent and s the estmated relatonshp between the return on the ndvdual asset, R t, and the excess return on the market portfolo, The market portfolo s theoretcally defned as the market value-weghted portfolo of all traded assets n the economy. Naturally, t s very dffcult to observe and measure the return of the true market portfolo n practce. Researchers therefore often use the returns of a broad stock market ndex as a proxy for the market portfolo even f ths s theoretcally ncorrect. Eugene F. Fama and Kenneth R. French (199) 16, found that beta was cross-sectonally statstcally nsgnfcant and proposed a three factor model wth the followng specfcaton: R mt. Rt Rmt SMBSMB HML HML t (Eq.)) Where Rt s the excess return of asset, at tmet, s a regresson coeffcent, s as before the estmated relatonshp between R mt R t and the excess return on the market portfolo,. SMB s the return on a factor mmckng portfolo constructed from small company returns mnus bg company returns. HML s the return on a factor mmckng portfolo 16 Fama, E and K.R French, 199, The cross secton of expected stock returns. 13
constructed from companes wth hgh book-to market ratos mnus companes wth low book-to-market ratos, and SMB and HML are the factor loadngs for asset,, on the SMB and HML portfolos, respectvely. One can also consder the Arbtrage Prcng Theory. The APT normally ncludes several systematc rsk factors. For example Chen, Roll and Ross (1986) estmate the model: R t IP EI UI CG GB (Eq.3) IP t EI t UI t CG t GB t t Where IP s the percent change n ndustral producton, EI s the percent change n expected nflaton, UI s the percent change n unexpected nflaton, CG s the excess returns of long-term corporate bonds over long-term government bonds, and GB s the excess returns of long-term government bonds over T-blls. 17 The unexplaned part of the return s denoted as the resdual and s defned as: t R R (Eq.4) t mt Clearly, the hgher the explanatory power, measured as R of the asset prcng model, the smaller the unexplaned part of returns, and the hgher the possblty of detectng the event s effect on return. To llustrate ths, consder the case where the researcher use a poor performng asset prcng model wth a low R. If a large resdual s obtaned, t wll be dffcult to tell whether ths s due to the event or due to the poor performance of the asset prcng model. Thus, the goal should be to fnd an as good model as possble. However MacKnlay (1997) argues that the margnal explanatory power of addtonal factors to the market beta usually s qute low, mplyng that the gans of usng multfactor models are lmted. 18 The am of ths thess s not to fnd the perfect asset prcng model. Therefore, the CAPM s used for the event study analyss. 17 Bode, Z., A. Kane and A.J. Marcus, 005, Investments, Sxth Edton, McGraw-Hll, Sngapore, p. 47 18 MacKnlay, A. Crag, 1997, Event Studes n Economcs and Fnance. 14
3.3 Expected answers to stated hypotheses based on economc theory As mentoned above, the peckng order theory mples that SEO announcements n general are related to company overvaluaton. The answer to hypothess 1 s therefore expected to be yes, and that SEO announcements n general should be assocated wth negatve abnormal returns, at least over short event wndow horzons. From economc theory, the answer to hypothess s not perfectly clear. However, emprcal fndngs n both Eckbo and Masuls and Andersson and Söderberg suggest that the answer to hypothess also should be yes, that non-underwrtten SEOs have sgnfcantly less negatve returns than underwrtten ones. Regardng hypothess 3, we have not been able to fnd an exstng theory or emprcal work descrbng the ssue n the lterature. However, one could argue that a SEO announcement n general s a complex process wth many elements that can ncrease uncertanty about the company s future, and that ths uncertanty would lead to an ncrease n return varance around the event date. The answer to hypothess 3 s therefore expected to be yes. Regardng hypothess 4, we have not been able to fnd an exstng theory or emprcal work. However, as descrbed n the chapter Benefts of underwrtng, together wth the opton theory analogy, explaned n the problem dscusson, one could argue that underwrtng should contrbute to lowerng the event wndow varance compared to non-underwrtten SEOs, and that the answer to hypothess 4 therefore also should be expected to be yes. 15
4. Data and Varables 4.1 Data sample In ths event study, daly stock prce data for 5 companes performng SEOs, lsted on the OMX Stockholm Large-cap, Md-cap and Small-cap exchanges, s used. The stock prce data s gathered from Thomson Reuters Datastream. The sample perod s January 006 to December 010. Stocks lsted on smaller Swedsh exchanges are not ncluded, snce small companes often are subject to varous knds of frm specfc rsks, such as lqudty rsk and effects of non-synchronous tradng. 19 By only ncludng companes on larger lsts, companes wth hgh turnover velocty, spurous effects n return propertes caused by nfrequent tradng are mnmzed. Turnover velocty s defned as the rato between the Electronc Order Book (EOB) turnover of domestc shares and ther market captalzaton. 0 Below, a chart of the turnover velocty for the Swedsh OMX exchange s presented. Chart 1. Turnover velocty for the Swedsh OMX Exchange. 150% OMX Turnover velocty 007-01 100% 50% 0% Small cap EUR <50m MId cap EUR 150-1.0bn Large cap EUR >1.0bn Data: Nasdaq OMX The sample sze of underwrtten SEOs s 45 and the sample sze of non-underwrtten SEOs s 7. 19 The effects of non-synchronous tradng are well descrbed n Asgharan, 010. 0 http://www.world-exchanges.org/statstcs/statstcs-defntons/turnover-velocty 16
4. Delmtatons n the data sample No SEOs wth unts has been ncluded n the study. A unt s as a stock coupled wth a subscrpton opton. Unts are not studed snce ths method dffers from the tradtonal rghts ssue wth prmary preferental rghts, and also due to the fact that unts are a rather uncommon flotaton method. No SEOs coupled wth Greenshoe optons are ncluded, snce these agreements unnecessarly complcate the analyss. No IPOs are ncluded, due to the lack of avalable hstorcal stock prce data. No drected share ssues are ncluded, due to the fact that drected shares are not publcly traded n the stock exchange. 4.3 Varables 4.3.1 The rsk free nterest rate The rsk free nterest rate used n ths event study s the Swedsh 90-day SSVX rate. On a daly bass, ths rate s reported as a smple yearly nterest rate. That s, the nterest rate obtaned by buyng a 90-day SSVX four tmes wthout compoundng t. To use ths rsk free rate n our CAPM-model t s necessary to transform the yearly 90-day SSVX rate nto a daly rsk free nterest rate. The daly nterest rate, R s calculated as: f 1 / 360 R 1 r 1 (1) f yearly 4.3. Return varables Ths event study s carred out usng daly excess returns for ndvdual assets and the market portfolo, respectvely. The excess returns for the ndvdual assets, R are calculated as: R R () R f Where R s the observed daly return on asset, and The excess return on the market portfolo, R s calculated as: m R f s the rsk free nterest rate. R R (3) m R f 17
Where R s the daly market return and market portfolo we use the Stockholm OMXSPI Index. 4.4 Testng the varables for statonarty R f s the rsk free nterest rate. As proxy for the Before estmatng the regresson model, t s mportant to frst test both the dependent and explanatory varable for statonarty. Statonarty s mportant, because f one of the seres n an equaton s found to be non-statonary, we rsk estmatng a spurous regresson relatonshp. 1 A random varable s sad to be covarance-statonary, or weakly statonary, f t has the followng propertes: Ths mples that: E t 1. Y t Var t. Y t 0 Cov Y t, Y t t 3. h h 1. The mean functon should be constant and ndependent of tme. No tme-trend should be present n the data.. The varance should be fnte and constant throughout the sample. 3. The autocovarance functon should be ndependent of tme. The covarance should depend only on the tme lag, h and not the tme perod tself. For all seres n our data sample, the statonarty tests are carred out n EVews by runnng the Augmented Dckey-Fuller, ADF-test, wth the specfcaton: Y t a a t Yt Yt (Eq.5) 0 1 1 Where EVews s set to automatcally select the number of lagged values of the Schwarz nformaton crteron. The null hypothess for the ADF-test s: Yt based on 1 Informaton about spurous regressons can be found n Granger and Newbold, 1974, Spurous regressons n econometrcs, Journal of Econometrcs, p 117 18
H 0 Unt root H1 No unt root The t-statstcs and the correspondng p-values are presented n table 1 n the appendx. It s observed that the null hypothess s rejected for all seres n the sample. Thus, t s concluded that all dependent and ndependent varable seres are statonary and that we can proceed wth estmaton of the regresson model. 5. Method The emprcal research s carred out usng OLS-regressons based on the CAPM-model. Frst, the varables used n the regresson model are tested for statonarty usng the ADFtest procedure descrbed above. After confrmaton of statonarty, estmaton of the regresson model s performed. The Gauss Markov assumptons of a lnear regresson model are tested usng approprate econometrc tests. For each event wndow horzon, cumulatve abnormal returns, CARs, are calculated and aggregated over securtes. The aggregated average CAR for each horzon, CAR s then tested for sgnfcance usng a J-test. For the varance tests, t s frst tested whether there s a sgnfcant ncrease n the return varance durng the month followng the SEO announcement compared to the varance prevalng durng two months before the announcement. Ths s done wth a varance rato F-test. It s then nvestgated whether there s a sgnfcant dfference n event wndow varance between underwrtten SEOs and non-underwrtten SEOs. In order to perform ths analyss, the full sample of companes s sorted nto two portfolos, one wth underwrtten and one wth non-underwrtten SEOs. The varance rato between these two portfolos s tested for sgnfcance usng an F-test. 5.1. Event study methodology An event study conssts of an estmaton wndow, where model parameters are estmated, and an event wndow where the effect of the event s analyzed. As noted above, we follow the conventonal event study methodology outlned n MacKnlay (1997). To llustrate our event study methodology we present fgure 1. 19
Fgure 1: Event study methodology Estmaton wndow Event wndow -6-1 t 0 t 1 - Event date 5. Estmaton Wndow Our estmaton wndow conssts of 6 tradng days and s selected as one year before the frst date n the event wndow. For the 6 tradng days, the daly return for each ndvdual asset, and the correspondng daly market return are calculated accordng to the formulas: pt pmt Rt 1 and, R mt 1 pt 1 pmt 1 To obtan beta estmates for each company, the followng regresson s estmated: (4) Rt Rmt t (Eq.6) The result s 5 beta coeffcents, whch are all found to be postve and sgnfcant on the 5 percent sgnfcance level. The beta coeffcents are nserted n the model for normal returns, (Eq. 7). 5.3 Event Wndow Our event wndow s one month long and stretches from two weeks before to two weeks after the announcement date. For ths perod we calculate the ndvdual stock returns usng the formula 4 above, as well as the so called normal returns and abnormal returns. The event wndow horzons are one day, three, seven, 15 and 9 days long, respectvely. The three day event wndow s measured as the announcement date 1 tradng day. The seven day event wndow s measured as the event day 3 tradng days, and so forth. 5.4 Measurng normal returns The normal return s the return of a certan asset, that would have been expected f the event dd not take place,.e. the return under normal condtons, when the asset moves 0
wth the market. To calculate the normal return, R for asset at tme t, the market model s used: nt R nt mt R (Eq.7) where s the estmated beta coeffcent for a certan asset,, obtaned from regressons n the estmaton wndow and Rmt s the daly market return. 5.5 Measurng abnormal returns The abnormal return, CAPM model. AR t s defned as the return n excess of what s predcted by the ARt s calculated as: Where R s the observed daly return on asset, and t 5.6 Dagnostc checkng of the regresson model R nt s the normal return. After estmaton of the regresson model, t s oblgatory to perform dagnostc checkng of the resduals. It s needed to test f the Gauss Markov assumptons of a lnear regresson model hold. If the Gauss Markov assumptons are volated, the OLS estmator wll no longer be the best lnear unbased estmator (BLUE estmator) of. The OLS estmator wll stll be unbased, but not effcent. If OLS s not BLUE, the varables wll have to be transformed before applyng OLS agan. The Gauss Markov assumptons are: AR t R R (5) 0 E t t nt Heteroskedastcty: Var t Autocorrelaton: Cov, 0 j j The resduals are also tested for normalty usng the Jarque-Bera test. If the resduals are not normally dstrbuted, nference based on the standard F-tests and t-tests wll not be vald. 1
5.7 Testng for Heteroskedastcty From each of our orgnal regresson equatons the resduals are obtaned and stored. For each resdual seres,, tests for heteroskedastcty are performed usng Whte s test n EVews. Whte s test s chosen snce t has hgher power aganst a general structure of heteroskedastcty, whle the Breusch-Pagan test has hgher power when the structure of the heteroskedastcty s known. The Whte s test procedure s carred out by estmatng the followng auxlary regresson: The test statstc, N R s asymptotcally N s the sample sze, R R (Eq.8) 0 1 t p dstrbuted wth p degrees of freedom, where R s the coeffcent of determnaton from the auxlary regresson and p s the number of regressors n the auxlary regresson, excludng the constant. The null hypothess for Whte s test s: H 0 Homoskedastcty H1 Heteroskedastcty The results from the heteroskedastcty tests are presented n table 1 n the appendx. The null hypothess of homoskedastcty s rejected for 10 out of 5 companes. Ths s a problem, snce t ndcates that the regresson model s not satsfactorly specfed for these companes. 5.8 Testng for Autocorrelaton All resdual seres are tested for autocorrelaton usng the Breusch-Godfrey LM-test n Evews. The auxlary regresson s estmated accordng to: R (Eq.9) 0 1 1 1 3 3 4 4 5 5 6 6 t Murray, M.P, 006, Econometrcs, A modern ntroducton, Pearson Educaton Inc. 3 MacKnlay, Crag. A, 1997, Event Studes n Economcs and Fnance.
The resduals are regressed aganst a constant, the regressor and sx lagged values of the resduals. The test statstc, N R s asymptotcally p dstrbuted wth p degrees of freedom, where N s the sample sze, R s the coeffcent of determnaton from the auxlary regresson and p s the number of lagged resduals n the auxlary regresson. The null hypothess for the LM-test s: H 0 No autocorrelaton n the resduals H1 Autocorrelaton n the resduals The observed test statstcs and ther correspondng p-values are presented n table 1 n the appendx. The null hypothess of no autocorrelaton s rejected for 16 out of 5 companes. Ths ndcates that ether the dependent varable, the ndependent varable or both exhbt problems wth autocorrelaton. 5.9 Jarque-Bera normalty test The resduals are tested for normalty usng the Jarque-Bera test. The Jarque-Bera test s constructed to detect devatons from the normal dstrbuton and the test statstc s calculated as: S JB N * 6 K 3 4 Where N s the sample sze, S s the sample skewness and K s the sample kurtoss. The JB- test statstc s dstrbuted wth degrees of freedom. The null hypothess for the JB-test s: H1 H Data s normally dstrbuted 0 Data s not normally dstrbuted The test statstcs and ther correspondng p-values are presented n table 1 n the appendx. Accordng to the Jarque-Bera test, the null hypothess of normally dstrbuted resduals s rejected for 50 out of 5 companes. (6) 3
5.10 Correctng for heteroskedastcty and autocorrelaton For companes exhbtng ether heteroskedastcty or autocorrelaton, the regressons are re-estmated usng the opton of Newey-West heteroskedastcty and seral correlatonconsstent standard errors n EVews. Ths procedure changes the standard errors, and therefore the t-statstcs, qute sgnfcantly, however not enough to change the sgnfcance of the varables. The Newey-West standard errors support vald nference wth OLS. 5.11 Testng the sgnfcance of average CAR For each horzon, cumulatve abnormal returns are aggregated through securtes and an average CAR, CAR, s calculated as: CAR N 1 (7) N 1 CAR In order to make an nference about whether or not CAR s sgnfcantly dfferent from zero, t s necessary to calculate the varance ofcar. In practce, because the true varance of the ndvdual CARs, s unknown, an estmator must be used to calculate the CAR varance of the abnormal returns. The sample varance of CAR from the market model regresson n the estmaton wndow s an approprate choce. 3 The ndvdual resdual varances from the estmated market model regressons are aggregated and the varance of CAR s calculated as: Var N 1 CAR N 1 CAR (8) For each horzon, t s tested f CAR s sgnfcantly dfferent from zero usng the test statstc: J CAR Var CAR a ~ N(0,1) The J-test statstc s asymptotcally standard normal dstrbuted. (9) 3 MacKnlay, Crag. A, 1997, Event Studes n Economcs and Fnance. 4
5.1 Testng for equalty of sample means To test our hypothess, whether or not the returns n the portfolo of underwrtten SEOs are sgnfcantly dfferent from the portfolo of non-underwrtten SEOs, a t-test for equalty of sample means s performed. The test statstc s calculated as follows: t x s 1 1 n 1 x s n (10) Where x 1 and x are the two sample means, and n s1 and s are the two sample varances, and n1 are the two sample szes. Snce our sample szes are not large enough to be approxmated by the normal dstrbuton, the t-dstrbuton s used. Also, snce the populaton varances are unknown and not assumed to be equal, a common number of degrees of freedom for the test, has to be calculated as: 4 s n The null hypothess for ths test s: 1 1 s n 1 1 s n s /( n1 1) /( n 1) n (11) H 0 : x1 x 0 Versus the alternatve hypothess: H 1 : x1 x The decson rule s to reject H 0 f: t t, /, or f t t, / 0 4 Newbold, P., W.L. Carlsson and B. Thorne, 006, Statstcs for busness & Economcs, 6 th Edton, p.378. 5
5.13 Varance tests To answer our hypothess 3,.e. to determne whether a SEO announcement n general has an effect on the return varance of the ssung company, a varance rato test s conducted. The rato between the return varance two months before the event wndow versus the varance for the month followng the announcement day s studed. For all assets,, the varance rato, VR, s calculated as: VR S S x y ~ Fn x 1, n 1, y (1) where S x s the largest of the two sample varances. For ths F-test to be correctly executed, t s requred to organze the varance rato wth the larger varance n the numerator and the smaller varance n the denomnator. 5 The null hypothess of the F-test s: H S x S : 0 y Versus the alternatve hypothess: H S x S : 1 y It s thus a one-sded test. The decson rule s to reject H 0 f: S S x y F n x 1, n 1, y. To test hypothess 4, a varance rato test s agan performed. The test procedure s the same as descrbed for hypothess 3 but usng the samples underwrtten and nonunderwrtten SEOs. Due to the absence of event wndow clusterng, the portfolo varance can be calculated usng smple aggregaton as descrbed n Eq. 8. 6 5 Ibd, p. 391. 6 The calculaton of CAR varance when event wndows are clustered are descrbed n MacKnlay, 1997, Event Studes n Economcs and Fnance. 6
5.14 Relablty The SEO prospectuses were requested from the Swedsh Fnancal supervsory authorty, Fnansnspektonen. Snce the prospectuses are approved by the same authorty, errors due to ncorrect prospectus nformaton are unlkely. Relevant nformaton has been carefully extracted from the prospectuses. After nformaton extracton, our study s strctly quanttatve, whch would facltate a replcaton made by other researchers. Our event study methodology follows the conventonal standard n the lterature, whch s outlned n MacKnlay (1997). Snce our delmtaton process result n a sample solely contanng rghts ssues wth prmary preferental rghts, the rsk of makng analyss based on other flotaton methods s mnmzed. The fact that our sample perod covers a severe fnancal crss may lead to more negatve, and more volatle returns than what would be found by other researchers for a dfferent sample perod. 5.15 Valdty As wth any regresson coeffcent, the beta coeffcents estmated n the regresson model are measured wth error. Snce all beta coeffcents are stll found to be sgnfcant on the 5 percent sgnfcance level, ths should be consdered to be of mnor sgnfcance. Due to the relatvely strct delmtatons, total sample sze s farly small, 5 companes. The sample sze of underwrtten SEOs s 45, and the sample sze of non-underwrtten SEOs s only 7. The small sample sze of non-underwrtten SEOs decreases the valdty of nference for that sample. However, both our samples can be seen as extenson to the sample studed n Andersson and Söderberg, and n that context, ths problem becomes less severe. The event study estmatons are based on the CAPM-model only. The CAPM s a lnear model specfcaton and s thus restrcted to measurng a lnear relatonshp between the asset and the source of systematc rsk. It s possble that a more complex model can provde a better ft to the data. For example, usng the Fama-French three factor model or an APT model could lead to dfferent results. The CAPM-model mplctly assumes that stock returns are normally dstrbuted. Our ndvdual company returns are, as results show, found not to be normally dstrbuted, 7
whle the market returns are found to be normally dstrbuted. The assumpton of normally dstrbuted asset returns s prmarly mportant for the correctness of aggregated portfolo rsk measures. For regresson model estmaton purposes, used n ths thess, the assumptons of normalty s not necessary for the orgnal CAPM-equaton to be vald as a regresson model. The CAPM s stll the estmated relatonshp between the ndvdual asset and the market return. The valdty of the conducted F-tests s somewhat dmnshed due to the fact that F-tests n general are qute senstve to the assumpton of normalty 7, whch s a problem snce our data shows ndcatons of non-normalty. 7 Newbold, P., W.L. Carlsson and B. Thorne, 006, Statstcs for busness & Economcs, 6 th Edton, p. 390. 8
6. Results of the event study 6.1 Results regardng returns, full sample When studyng the full sample ncludng all SEOs, a sgnfcant negatve average CAR of -.5 percent and a negatve medan CAR of -3.5 percent are observed for horzons one and three. Ths s n lne wth several prevous studes examnng announcement effects of SEOs. In general, our average CARs are found to be slghtly more negatve than what s presented n Andersson and Söderberg (007). Ths result mght arse due to the fact that our sample perod ncludes a severe fnancal crss. It can also arse due to a smaller total sample sze. Our results are summarzed n table below. Table. Average and medan CAR, full sample. All SEO's Non underwrtten SEO's Underwrtten SEO's Horzon Average CAR Medan CAR Average CAR Medan CAR Average CAR Medan CAR 1-0,0597-0,03513 0,01133-0,0399-0,03177-0,0377 3-0,0479-0,0346 0,05147 0,0194-0,03666-0,05354 7-0,01781-0,0499 0,0478 0,094-0,0793-0,0315 15-0,01641-0,0551 0,00658-0,031-0,01998-0,01881 9-0,01687-0,00841-0,0478 0,09770-0,0146-0,00841 The focus s preferred to be on medan CAR, snce the data sample contans large postve and negatve outlers n the CARs. However, to test for the sgnfcance of abnormal returns usng the J-test procedure descrbed above, one has to use average CAR,CAR. The results from the sgnfcance tests are presented n table 3 below. Table 3. Sgnfcance tests of average CAR. Test f average CAR s sgnfcant Horzon Average CAR Standard devaton J-test statstc p-value 1-0,0597 0,00490-5,30111 0,00000 3-0,0479 0,00849 -,9177 0,00559 7-0,01781 0,0196-1,37391 0,1555 15-0,01641 0,01897-0,8646 0,745 9-0,01687 0,0638-0,63941 0,3519 For horzon one and three, CAR s found to be sgnfcant on all commonly used sgnfcance levels. For horzons seven, 15 and 9, CAR s found to be not sgnfcantly dfferent from zero on all commonly used sgnfcance levels. Ths s due to the larger standard devaton n 9
returns for the longer horzons. It s thus notced, that for longer horzons, t becomes more dffcult to dstngush the mpact of the event. Ths s n lne wth the effcent market hypothess, whch states that new nformaton rapdly should be ncorporated n the stock prce. 6. Results regardng returns, sub sample When studyng the returns n the groups of underwrtten and non-underwrtten SEOs separately, t s observed that the average CAR for non-underwrtten SEOs s hgher than for underwrtten SEOs for all horzons except for CAR9. The medan CAR for nonunderwrtten SEOs s hgher for all horzons except for CAR15. The dfference n returns between the two groups s tested for statstcal sgnfcance usng the test procedure outlned n (10). The results are presented n table 4 below. Table 4. Test for the dfference of sample means. Horzon x1-x denomnator t-statstc p-value 1-0,04310 0,0165 -,6518 0,0917 3-0,0881 0,0139-4,1005 0,00334 7-0,0751 0,0644 -,84515 0,0164 15-0,0656 0,03198-0,83049 0,43035 9 0,03031 0,03771 0,80375 0,44477 It s observed that there s a sgnfcant dfference n returns for horzons one, three and seven, whle the dfference for horzons 15 and 9 s found to be nsgnfcant. 6.3 Results regardng varance, full sample For 40 out of 5 companes, the event wndow varance s found to be hgher than the estmaton wndow varance. 30 of these 40 varance ratos are found to be sgnfcant. For four companes the estmaton wndow varance s larger than event wndow varance. For 18 companes, the varance rato s not large enough to be sgnfcant on the fve percent level. Both the estmaton and event wndow varances for ndvdual companes are lkely to be affected by the overall market varance prevalng n each tme perod, respectvely. Below, a chart of the market returns for the OMXSPI ndex s presented. An ncrease n return 30
varance s observed n years 008 and 009, a perod when the global fnancal crss ncreased uncertanty n the fnancal markets. OMXSPI daly return varance 0,1 0,08 0,06 0,04 0,0 0-0,0-0,04-0,06-0,08-0,1 006-01-0 007-01-0 008-01-0 009-01-0 010-01-0 011-01-0 The majorty of the observatons wth hgher event wndow varance are located n the perod from February 008 to June 009, a perod wth ncreasng market return varance. For the four companes wth lower estmaton wndow varance, a very clear trend can be observed. These observatons are located between July 009 and September 009, a perod characterzed by decreasng market return varance. 6.4 Results regardng varance, sub samples The comparson between the 9 day event wndow varance for the sample of underwrtten SEOs and the 9 day event wndow varance for non-underwrtten SEOs s consdered below. Table 5. Test for sgnfcant dfference n return varance. Event wndow varance Varance F-rato P-value Underwrtten SEO's 0,003390968 1,499381438 0,30139381 Non-underwrtten SEO's 0,0061578 It s observed that the event wndow varance rato s not statstcally sgnfcant on all commonly used sgnfcance levels. The varance for the portfolo of non-underwrtten SEOs s lower than the varance for underwrtten SEOs. 31
7. Analyss and dscusson 7.1 Answers to hypotheses 1-4. H1: The answer to hypothess 1 s yes for short horzons up to three days, and no for horzons of seven days and longer. There s a sgnfcant negatve announcement effect on the event date. H: The answer to Hypothess s yes. The dfference s sgnfcant for horzons one, three and seven, whle nsgnfcant for horzon 15 and 9. H3: The answer to hypothess 3 s yes. There s n general an ncrease n return varance for the ssung companes durng the month followng the SEO announcement. H4: The answer to hypothess 4 s no. In the studed sample, there s no ndcaton that underwrtten SEOs have lower event wndow return varance than non-underwrtten SEOs. 7. Results n lne wth theory? The result n hypothess 1 s n lne wth the peckng order theory developed by Myers and Majluf (1984). The result n hypothess s n lne wth emprcal fndngs n both Eckbo and Masuls (199) and Andersson and Söderberg (007). Eckbo and Masuls s theory, that a non-underwrtten SEO s a sgnal of relatve company qualty, seems to better explan the results than the theory of underwrters beng certfers of value. The results n hypothess 3 provde support for the argument that a SEO s an event that ncreases uncertanty and thereby return varance. The results n hypothess 4 contradct our theoretcal expectaton that underwrtten SEOs should have lower event wndow varance than non-underwrtten ones. 7.3 Concludng dscusson Our results do not ndcate that underwrtten SEOs have better return propertes than nonunderwrtten SEOs. On the contrary, non-underwrtten SEOs seems to perform better than underwrtten SEOs, snce they are assocated wth less negatve returns coupled wth a lower varance. On a purely return based perspectve, one can therefore not see any clear benefts of underwrtng. 3
It s problematc, however, to make the causal statement that the nferor return propertes of underwrtten SEOs are due to the underwrtng tself, and not some other underlyng frm specfc varable or property not ncluded n the analyss. Snce no deep analyss tryng to determne or fnd support of causalty s performed, ths thess cannot conclude that the nferor return propertes are drectly caused by the underwrtng. It s merely an observaton, that n the studed sample, the group of underwrtten SEOs exhbt worse performance. Underwrtng stll has a purpose and functon for ssung companes despte that underwrtten SEOs, n our sample, the sample n Eckbo and Masuls as well as n the sample studed by Andersson and Söderberg, fal to show any advantages from a return based perspectve. Our fndngs are n lne wth the ones n Andersson and Söderberg (007), that Swedsh nvestors seem to devote lttle mportance to whether a SEO s underwrtten or not. Ths mght be because of varous reasons of whch some are hypotheszed below. Durng the last 10 years, underwrtten SEOs have become more or less standard. The market mght therefore no longer dstngush whether a SEO s underwrtten or not. The underwrter oblgatons are typcally not secured by any collateral, whch s mparng the qualty of the underwrtng. The nsecurty about whether the underwrter agreement s legally bndng or not s also mparng the qualty of underwrtng. Based on the fndngs of our event study analyss, we could suggest a decreased dependence on underwrters, a downward pressure on underwrter compensaton, or makng the underwrter compensaton condtonal on the ex post actual subscrpton. However, wthout consderaton of other research, the results from our thess should not alone be used as gudance for decsons. It s mportant to note that all matters regardng underwrtng are ultmately up to legslators and the market to decde. 33
8. References Artcles Andersson, M.E. and S. Söderberg, 007, Rghts Issues n the Swedsh Market, A Comparson between Insured and Unnsured Rghts Issues. http://arc.hhs.se/download.aspx?medumid=330 accessed 013-05-13 Hoffman Bermejo, W and R. Raudsepp, 009, En stude av problematken vd nyemssoner. http://arc.hhs.se/download.aspx?medumid=77 accessed 013-05-13 Prawtz, D., 009, Emssonsgaranter ur ett rättslgt perspektv. http://lup.lub.lu.se/luur/download?func=downloadfle&recordoid=1561351&fleoid=1565608 Accessed 013-05-13 Malmström, K. and A. Nlsson, 00, Annonserngseffekt av nyemssoner - En fallstude på Stockholmsbörsen http://www.lundunversty.lu.se/o.o..s?d=4965&postd=1340378 Accessed 013-05-13 Von Arronet, C., J. Källstrand, and M. Tarnawsk-Berln, 003, Kursreaktoner på tllkännagvande av nyemsson. http://www.lundunversty.lu.se/o.o..s?d=4965&postd=134998 Accessed 013-05-13 Frtzell, M. and J. Hansveden, 006, Stock Market Reactons and Offerng Dscounts of Swedsh Equty Issues. http://www.uppsatser.se/uppsats/76eac50e/ Accessed 013-05-13 Egerot, R., E Hagman, and M. Svensson, 009, Deltagande I nyemsson - en buy and holdstrateg. http://www.lundunversty.lu.se/o.o..s?d=4965&postd=1549355 Accessed 013-05-13 Gustavsson, M and P. Lndström, 010, Garanter vd nyemssoner förutsättnngar och kostnader. http://www.uppsatser.se/uppsats/1efb4ee8a/ Accessed 013-05-13 34
Månsson, M. and C. Rostedt, 010, Varnng för ras En stude av aktemarknadens reakton på nyemssonsbeskedet. https://gupea.ub.gu.se/handle/077/501 Accessed 013-05-13 Myers, S.C and N.S. Majluf, 1984, Corporate Fnancng and Investment Decsons When Frms Have Informaton That Investors Do Not Have, Journal of Fnancal Economcs, 13, 187-1. Eckbo, B.E. and R.W. Masuls, 199, Adverse selecton and the rghts offer paradox. Journal of Fnancal Economcs, 3, 93-33. Fama, E and K.R French, 199, The cross secton of expected stock returns, Journal of Fnance, Volume 47, Issue (Jun.199), 47-465. MacKnlay, A. Crag, 1997, Event Studes n Economcs and Fnance, Journal of economc lterature, Vol. XXXV (March 1997), pp. 13 39 Granger, C.W.J. and P. Newbold, 1974, Spurous regressons n econometrcs, Journal of Econometrcs, p 117 Books: Bode, Z., A. Kane and A.J. Marcus, 005, Investments, Sxth Edton. McGraw-Hll, Sngapore Murray, M.P, 006, Econometrcs, a modern ntroducton, Pearson Educaton Inc. Newbold, P., W.L. Carlsson and B. Thorne, 006, Statstcs for busness & Economcs, 6 th Edton. Prentce Hall, New Jersey. 35
Internet sources: Veckans affärer, nyemssoner onödgt dyrt, publshed onlne 009-03-6, Accessed 013-05-13 http://www.va.se/nyheter/nyemssoner-onodgt-dyrt-53775 Asgharan, H., 010, Lecture notes n Emprcal Fnance, page 9-33, Accessed 013-05-13 http://www.nek.lu.se/nekhas/documents/lecture%0notes%010.pdf Emssonsgaranter krtseras, publshed onlne 010-07-19, Accessed 013-05-13 http://www.aktespararna.se/artkelarkv/opnon/010/jul/emssonsgaranter-krtseras/ Defnton of turnover velocty: Accessed 013-05-13 http://www.world-exchanges.org/statstcs/statstcs-defntons/turnover-velocty Data source for turnover velocty: Accessed 013-05-30 http://nordc.nasdaqomxtrader.com/newsstatstcs/statstcsandreports/nordc_reports/ 36
9. Appendx Table 1. Results of statonarty tests and resdual dagnostc tests. Statonarty tests Resdual dagnostcs tests Y-varables X-varables ADF test ADF test Whte's test LM test Jarque-Bera test Company test, Stat p-value test, Stat p-value test, Stat p-value test, Stat p-value test, Stat p-value 1-17,1749 0,00000-7,54907 0,00000 8,89388 0,01170 5,4670 0,48600 98,54457 0,00000-15,0404 0,00000-7,53886 0,00000 0,1896 0,90950 8,44601 0,070 454,0000 0,00000 3-17,435 0,00000-7,7100 0,00000,9606 0,760 13,60355 0,03440 51,6898 0,00000 4-15,5800 0,00000-7,8757 0,00000 3,58365 0,16670 6,0775 0,41460 103,0750 0,00000 5-17,37565 0,00000-16,3519 0,00000,71169 0,5770 7,5543 0,7500 144,6740 0,00000 6-13,31898 0,00000-13,460 0,00000 0,41936 0,81080 10,03079 0,1340 99539,99000 0,00000 7-19,3140 0,00000-16,49177 0,00000 0,46856 0,79110 16,0031 0,01360 4,713 0,0943 8-17,8081 0,00000-18,13113 0,00000 7,894 0,01930 7,30073 0,9390 9,1099 0,0105 9-14,3389 0,00000-17,90355 0,00000 0,95934 0,61900 15,49655 0,01670 19477,11000 0,00000 10-17,9378 0,00000-18,33084 0,00000 0,56363 0,75440,53536 0,86450 4187,5000 0,00000 11-18,9791 0,00000-18,13971 0,00000 0,88 0,64330 11,4818 0,07460 36,7981 0,00000 1-16,58035 0,00000-17,8156 0,00000 0,8367 0,65950 5,395 0,50310 807,7010 0,00000 13-17,69596 0,00000-18,37830 0,00000 1,0977 0,54610 5,4019 0,49340 1535,68000 0,00000 14-7,0581 0,00000-17,33349 0,00000 4,0436 0,13370 3,99988 0,00000 4058,38900 0,00000 15-16,19099 0,00000-16,07569 0,00000 16,69936 0,0000 1,76404 0,04690 369,6370 0,00000 16-15,34906 0,00000-16,03335 0,00000 1,36470 0,50540 7,3055 0,9350 38,33084 0,00000 17-13,35954 0,00000-15,88696 0,00000 5,17455 0,00000 7,6981 0,6650 10,4807 0,00530 18-15,4644 0,00000-15,88696 0,00000 6,6561 0,03590 3,56038 0,73590 195,1160 0,00000 19-19,46074 0,00000-15,509 0,00000 3,16851 0,0510 15,1396 0,01930 6,77460 0,00000 0-14,8410 0,00000-15,56178 0,00000 0,80618 0,66830 7,6817 0,640 1914,30500 0,00000 1-16,1090 0,00000-15,536 0,00000 4,0144 0,140 6,10593 0,41140 041,88300 0,00000-14,10081 0,00000-15,63904 0,00000 10,49431 0,00530 9,3934 0,15590 6,1957 0,04461 3-15,57538 0,00000-15,63904 0,00000 0,5749 0,75110 8,77108 0,18690 590,0440 0,00000 4-1,86633 0,00000-15,5600 0,00000 0,8303 0,6600 1,534 0,05650 701,90000 0,00000 5-13,0067 0,00000-15,6011 0,00000 0,01431 0,9990 17,06554 0,00900 50,1715 0,00000 6-16,7088 0,00000-15,63084 0,00000 3,99800 0,13550 6,38849 0,38110 6,67533 0,0355 7-16,48134 0,00000-15,55936 0,00000 4,03773 0,1380 5,4968 0,48180 5,9539 0,05168 8-13,4553 0,00000-15,578 0,00000 0,6530 0,7140 1,00451 0,06190 48,08680 0,00000 9-15,35867 0,00000-15,57087 0,00000 10,1140 0,00630 7,69030 0,6170 688,730 0,00000 30-16,04335 0,00000-1,83800 0,00000 1,96905 0,37360 3,8189 0,7010 343,88690 0,00000 31-19,99514 0,00000-1,50734 0,00000 11,95478 0,0050 8,90608 0,00010 1514,80800 0,00000 3-19,4853 0,00000-15,43956 0,00000 1,58548 0,4560 17,48070 0,00770 11,84570 0,00000 33-14,8370 0,00000-1,50734 0,00000 0,43474 0,80460 3,09008 0,79750 684,65000 0,00000 34-13,936 0,00000-1,50734 0,00000 1,7374 0,00000 9,84333 0,13140 14,6896 0,00065 35-13,0645 0,00000-15,41343 0,00000 8,69064 0,01300 6,50530 0,36900 04,65700 0,00000 36-16,67347 0,00000-15,41343 0,00000 1,60538 0,44810 10,65013 0,09980 404,85750 0,00000 37-18,71350 0,00000-16,41590 0,00000 0,1358 0,93590 16,34034 0,0100 0,6460 0,00000 38-15,37469 0,00000-16,8191 0,00000 0,9953 0,60800 7,31399 0,980 1555,34800 0,00000 39-19,4760 0,00000-16,47436 0,00000 1,34439 0,51060 5,8315 0,0000 571,70 0,00000 40-13,90650 0,00000-16,6056 0,00000 1,86788 0,39300 17,1561 0,00880 556,56600 0,00000 41-18,48700 0,00000-16,6056 0,00000,38696 0,3030 5,87909 0,43690 303,7430 0,00000 4-14,0787 0,00000-17,01464 0,00000 0,8188 0,6640 13,8656 0,03870 94,43690 0,00000 43-17,78515 0,00000-17,01904 0,00000 0,1454 0,89830 6,41041 0,37880 369,87880 0,00000 44-16,63069 0,00000-16,3947 0,00000 3,09714 0,160,10809 0,90950 49,79950 0,00000 45-0,5677 0,00000-16,67079 0,00000 0,056 0,98730 15,41770 0,0170 41,60950 0,00000 46-16,89607 0,00000-16,38190 0,00000 0,31649 0,85360 5,47041 0,48500 018,84000 0,00000 47-17,40089 0,00000-16,38190 0,00000 1,8080 0,40490 15,846 0,01460 613,6700 0,00000 48-18,51176 0,00000-16,31446 0,00000 0,16874 0,91910 15,64998 0,01580 56,1456 0,00000 49-14,96081 0,00000-16,58514 0,00000 1,65086 0,43800 9,33005 0,15580 08,99330 0,00000 50-17,36673 0,00000-16,63086 0,00000 0,691 0,87410 4,35965 0,6810 13388,6000 0,00000 51-15,17500 0,00000-16,63086 0,00000 3,18 0,0990 6,1855 0,4080 117,30810 0,00000 5-14,35466 0,00000-16,80446 0,00000 0,6065 0,73980 6,57050 0,3640 7773,06100 0,00000 37