Information Content of Dividends: Evidence from Istanbul Stock Exchange

Transcription

1 Informaion Conen of Dividends: Evidence from Isanbul Sock Exchange Ayse Aliok-Yilmaz (Corresponding auhor) Depermen of Managemen, Bogazici Universiy 34342, Bebek-Isanbul, Turkey Tel: Elif Akben-Selcuk Depermen of Managemen, Bogazici Universiy 34342, Bebek-Isanbul, Turkey Tel: Absrac This sudy invesigaes he marke reacion o dividend change announcemens a he Isanbul Sock Exchange. A sample of 184 announcemens made by 46 companies during he period 2005 o 2008 is analyzed by using he even sudy mehodology. The resuls sugges ha he marke reacs posiively o dividend increases, negaively o dividend decreases and does no reac when dividends are no changed, consisen wih he signaling hypohesis. Also, he resuls show pre-even informaion leakage for he decreasing dividends sample. Keywords: Dividends, Informaion Conen, Signaling Hypohesis, Even Sudy, Isanbul Sock Exchange 1. Inroducion A grea deal of work has been done in he areas of marke reacion o dividend announcemens and he informaion conen of dividends hypohesis, which saes ha dividends convey o he marke paricipans informaion abou fuure cash flows of a company. However, previous empirical sudies of dividend signaling and informaion conens of dividends produced mixed resuls. Hence, a greaer undersanding of he issue is warraned. This sudy conribues o he exising lieraure by esing he informaion conen of dividends hypohesis for an emerging marke, namely he Isanbul Sock Exchange (ISE). ISE has been founded in 1986 and played an imporan role in he financial developmen in Turkey. This sudy covers he period 2005 o 2008, and invesigaes 184 dividend announcemens made by 46 seleced companies. Specifically, he impac of posiive, negaive or consan dividend announcemens on sock prices is invesigaed. The remainder of he aricle is organized as follows. Secion 2 presens he review of lieraure. The daa and research mehodology are described in secion 3. Secion 4 discusses he empirical resuls. Finally, secion 5 summarizes he main findings of he sudy and concludes. 2. Lieraure Review In a world where relevan informaion abou he firm s fuure prospecs is rare and cosly, dividends are considered an imporan source of informaion for invesors. The informaion conen of dividends heory indicaes ha managers use dividend announcemens o signal heir beliefs abou he prospecs of he firm. Since managers spend mos of heir ime analyzing he firm s operaions, sraegies and invesmen opporuniies, hey have beer and imelier informaion abou he firm s performance. Therefore, an announcemen of an increase in he dividend rae reflecs he managemen s view ha he firm s fuure earnings and cash flows are expeced o rise. An increase in dividends conveys good informaion abou a firm because i proves ha a firm is able o generae cash. Similarly if a company announces a decrease in dividend, he invesors would perceive i as a negaive signal regarding he curren and fuure performance. The informaion conen of dividends is also called signaling effec of dividends. Linner was he firs researcher o employ he erm informaion conen of dividends. In his sudy in 1956, he suggesed ha he managers only increase dividends when hey believe ha earnings of he firm have permanenly increased. Laer in 1961, Miller and Modigliani also discussed he informaion conen of dividends. According o heir discussion of dividend policy under uncerainy, in he real world, a dividend rae change is ofen followed by a 126

2 change in he sock price change. Bu hey argue ha such a phenomenon would no be incompaible wih heir irrelevance proposiion o he exen ha i was merely a reflecion of wha migh be called he informaional conen of dividends. They explained he informaion conen of dividends as follows. If a firm has a generally appreciaed arge payou raio and a sable dividend policy, invesors are likely o inerpre a change in he dividend rae as a change in managemen s beliefs of fuure profi prospecs for he firm. One of he firs empirical sudies abou he signaling effec of dividends was published in 1969 by Fama, Fisher, Jensen and Roll. In heir paper, hey examined wheher and how common sock prices adjus o he informaion involved in a sock spli. They hypohesized ha sock splis migh be inerpreed as a message abou dividend increases. The fac ha he cumulaive average residuals for boh dividend classes (increasing and decreasing) rise sharply in he few monhs before he spli, is consisen wih he hypohesis ha he marke recognizes ha splis are usually associaed wih higher dividend paymens. Their resul also suppors ha he sock marke is efficien in he sense ha sock prices adjus very rapidly o new informaion. In anoher paper, Pei (1972) aemped o offer furher evidence abou he validiy of he efficien marke hypohesis by esimaing he speed and accuracy wih which marke prices reac o announcemens of changes in dividends. His resuls suppor he argumen ha marke paricipans make considerable use of he informaion implici in announcemens of changes in dividend paymens. He indicaed ha he marke reacs very quickly o he announcemens when dividends are reduced or increased subsanially bu he effec of a moderae increase or decrease is relaively less. He also menioned ha wih few excepions, larges single effec occurs in he announcemen monh and he marke is reasonably efficien on boh a monhly and daily basis. He concluded ha he announcemen of dividend changes convey significan informaion. Number of empirical sudies followed he work of Pei (1972). In one of hem, Was (1973) examined he hypohesis ha dividends conain informaion abou he fuure earnings of he company and ha knowledge of pas and curren dividends enables a good predicion abou fuure earnings. He concluded ha he poenial informaion in dividends is very small. He concluded ha he informaion in dividends could only be rivial. In anoher sudy, Chares (1978) aimed o assess marke efficiency wih respec o dividend informaion and o documen risk and reurn behavior of socks around dividend changes by using monhly price daa and quarerly dividend informaion. He found significan abnormal reurns in monhs following he announcemens of dividend changes. His invesigaions wih daily daa also suppored his monhly resuls. Bu he concluded ha daily evidence does no necessarily reveal he presence of informaion in dividend announcemens since he made no effor o remove he effec of conemporaneous earning announcemens. Bhaarcharya (1979) assumed ha ouside invesors have imperfec informaion abou a firm s profiabiliy and ha cash dividends are axed a a higher rae han capial gains. His findings showed ha despie he ax disadvanage of paying dividends, firms migh pay dividends because dividends funcion as a signal of expeced cash flows. Aharony and Swary (1980) aimed o verify wheher quarerly dividend changes convey informaion beyond ha already provided by he earnings number. Their findings indicaed ha sockholders of firms which did no change heir dividends earned normal reurns during he even period. The abnormal reurns were no significanly differen from zero. In he case of dividend increases, sockholders earned posiive abnormal reurns, and mos of he saisically significan abnormal reurns occurred during days 1 and 0. In case of dividend decreases, sockholders had negaive abnormal reurns during he weny days surrounding announcemen daes, and similar o he case of dividend increases, mos of he significan abnormal reurns occurred during days 1 and 0. They concluded ha changes in quarerly cash dividends do provide informaion abou changes in managemen s assessmen of firm s fuure performance, so hey suppored he informaion conen of dividend hypohesis. In anoher sudy, Asquih and Mullins (1983) invesigaed he impac of dividends on sockholders wealh by analyzing 168 firms ha paid no dividends eiher during heir corporae hisories or for a leas he las en years. They used daily daa, and esimaed abnormal reurns for each firm around he announcemen daes. Then hey calculaed he average excess reurns and average cumulaive excess reurns. They also calculaed he 2-day average excess reurn for each dividend announcemen hey examined, so as o capure he enire impac of a dividend announcemen. Their resuls for he wo-day announcemen period indicaed significan and large posiive excess reurns. Their resuls are several imes larger han Chares s (1978) and Aharony and Swary s (1980) resuls. In anoher empirical research, Kalay and Loewensein (1985) had he purpose o demonsrae ha radiionally measured excess reurns over an even period, could reflec he higher compensaion ha risk averse invesors Published by Canadian Cener of Science and Educaion 127

3 require o hold he asse over a riskier period. They chose dividend announcemens as he even and esed he hypohesis ha he mean reurns on days in he even period are no significanly differen from he means on any oher random day. They consruced a sample of 302 dividend announcemens and esed heir hypohesis by using boh he marke model and mean adjused reurns model o esimae he abnormal reurns. Their resuls showed ha he mean excess reurns over he even period were significanly higher han hose on a random day. Their suggesed explanaion for hese higher reurns was higher risk per uni of ime in he even period. Michaely, Thaler, and Womack (1995) invesigaed he immediae and long-erm marke reacions o iniiaions and omissions of cash dividend paymens. Their sample conained 561 cash dividend iniiaions and 887 cash dividend omissions beween 1964 and By using buy and hold mehod hey calculaed he excess reurns of he securiies for a hree-day even period, and for monhly periods before or afer he even respecively. They found ha immediae impac of dividend omissions is negaive and immediae impac of dividend iniiaions is posiive. Their findings saed ha he sock prices coninued o rise even afer he iniiaion announcemen. For he omissions, here is a drif in he negaive direcion. They also noed ha he long-erm resuls of he omission sample are more robus han hose of he iniiaion sample. In 1997, Benarzi, Michaely, and Thaler examined wheher dividend changes give informaion abou fuure earning changes. They argued ha alhough here is much evidence abou he marke s response o dividend changes as newsworhy, here is less knowledge abou he acual realizaion of fuure earnings. By aking 1025 firms and 7186 dividend announcemens beween years as a sample, hey ried o deermine wheher changes in dividends have informaion conen abou fuure earnings. Their empirical resuls were consisen wih Wa s findings in They could no find much evidence of a posiive relaionship beween dividend changes and fuure earnings changes. They also invesigaed Linner s (1956) argumen ha dividend increases are a signal abou a permanen shif in earnings raher han a signal abou fuure earnings growh, and hey found a srong pas and concurren link beween earnings and dividend changes. They concluded ha changes in dividends mosly ell somehing abou wha has happened and ha Linner s model of dividends remains he bes descripion of he dividend seing process available. Dyl and Weigand (1998) invesigaed he changes in firm risk following he iniiaion of cash dividends. They inroduced he risk-informaion hypohesis ha dividend iniiaion conveys informaion o he marke abou he firm s lower risk. They proposed ha managemen s decision o iniiae dividend paymens per se provides he marke wih new informaion regarding he risk of he firm, and firm risk will be lower following dividend iniiaion. They esed heir hypohesis and invesigaed wheher he announcemen of dividend paymens is accompanied by a change in sysemaic risk by comparing firms pre and pos announcemen beas by using he Fowler-Rorke mehodology. Their findings indicaed ha boh he oal risk and he sysemaic risk of a sample firms ha iniiaed dividends are significanly lower in he year following he dividend announcemen. Garre and Priesly (2000) analyzed he dividend behavior of he aggregae sock marke. They proposed a model ha assumes managers minimize he coss of adjusmen associaed wih being away from heir arge dividend payou and hey found significan evidence of dividends conveying informaion regarding unexpeced posiive changes in curren permanen earnings. In 2001, Fama and French presened he phenomenon of he disappearing dividend. They found ha given heir characerisics, firms become less likely o pay dividends. They used logi regressions and a porfolio approach o documen ha characerisics and propensiy o pay make large separae conribuions o he decline in he percen of payers. Gunasekarage and Power (2002) reexamined he dividend-signaling hypohesis by examining he pos-announcemen performance of UK companies ha disclose dividend and earnings news o he marke on he same day. Their sample consised of 1787 earnings and dividends changes announcemens made on he same day beween he years They esed he hypohesis ha dividend-increasing companies do no ouperform heir dividend decreasing counerpars during he pos announcemen years wih respec o financial performance. By using he buy and hold mehod hey calculaed he abnormal reurns for he year before he announcemen, for he quarer before he announcemen and for over he announcemen period. Their findings did no provide a suppor for he signaling hypohesis. In one a more recen paper, Grullon e al. (2005), conrary o he dividend signaling hypohesis, showed ha dividend changes are uncorrelaed wih fuure earnings changes when conrolling for he well-known nonlineariies in he earnings process. Amihud and Li (2006) proposed an explanaion for he disappearing dividend phenomenon which was presened by Fama and French in They found ha a reason for a decline in he informaion conen of 128

4 dividends is he rise in holdings by insiuional invesors ha are more sophisicaed and informed. They found a decline in abnormal reurns around dividend change announcemens since he mid-1970s. Across firms, abnormal reurns are a decreasing funcion of insiuional holdings. They argued ha insiuional invesors exploi heir superior informaion and buy before dividend increases and ha dividends are less likely o rise in firms wih high insiuional holdings. Recenly, Yılmaz and Gunay (2006) examined he effecs of cash dividend paymens on sock reurns and rading volumes in he Isanbul Sock Exchange from 1995 o They found ha prices sar o rise a few sessions before cash dividend paymens, and fall less han dividend paymens on he ex-dividend day, finally decreasing in he sessions following he paymen. The resuls of rading volume analysis showed a considerable upward shif before he paymen dae and ha he volume became sable afer he ex-dividend dae. The findings suppored price-volume reacion discussions on he dividend paymen dae and he significan effec of cash dividends on he sock marke. As i becomes clear from he preceding discussion, previous empirical sudies of dividend signaling and informaion conens of dividends produced mixed resuls. 3. Daa and Mehodology The objecive of his sudy is o invesigae wheher dividends have informaional conen. To accomplish his objecive, an even sudy is conduced o measure wheher any abnormal reurns are earned by securiy holders around dividend announcemens. The core assumpion of he even sudy mehodology is ha if informaion communicaed o he marke conains any useful and surprising conen, an abnormal reurn will occur. A he ime of he even, he magniude of he abnormal performance indicaes he impac of ha paricular even on shareholders wealh. Following MacKinlay (1997), he firs sep of he analysis was o deermine he sample of firms o be included in he analysis. For he purposes of his sudy, ISE-lised companies ha regularly disribued cash dividends for he period 2005 o 2008 were seleced. In order o miigae he effec of oher conemporaneous evens on sock prices, any company wih an announcemen relaed o earnings, sock splis or merger and acquisiions around he dividend announcemen were excluded from he analysis. Financial secor firms were also excluded as hey are subjec o differen regulaions, resuling in a final sample of 46 firms and 184 dividend announcemens. The dividend announcemen daes and amouns of dividends paid by hese companies were colleced from he daily bulleins of ISE. The dividend announcemen dae (day 0) was aken as he day on which dividend amoun of a firm is announced o he public in he daily bullein. Neiher he ex-dividend day nor he day he dividend is paid was considered o be he announcemen day. As he second sep of he even sudy, a hree-day symmeric even window was chosen (1 day prior o he even day and 1 day afer he even day). This window lengh is appropriae o capure any news ha migh have leaked shorly before he official announcemen was made and also considers any shor-erm sock price reacions linked o he even afer he announcemen (Kohari and Warner, 2004). In addiion, several oher window lenghs are analyzed o check he resuls. The hird sep was he predicion of a normal reurn during he even window in he absence of he even. To esimae normal reurns, his sudy uses 355 daily reurns as he esimaion period, from he day T=-360 o he day T=-6.The model used in his sudy o esimae he expeced reurns is he marke model. I is a linear ime-series model where dependen variable, securiy reurns, is regressed agains percenage changes in a marke index. The marke model used in his sudy for securiy i for he year j during period can be expressed by he following linear ime-series model. R i, j, = α i, j + βi, j, Rm, j, + ei, j, where; R i,j, = daily reurn on he securiy i for year j during ime α i,j, β i,j = marke model parameers for securiy i for year j, securiy-specific inercep and slope coefficiens R m,j, = reurn of he marke (ISE-100 index) for year j during ime e i,j, = error erm for securiy i for year j a period. I is assumed ha e i,j, fulfills he assumpions of he linear regression model. Namely e i,j, has he mean of zero over he regression period, and has a variance independen over ime. The fourh sep was he calculaion of he abnormal reurn wihin he even window, where he abnormal reurn is defined as he difference beween he acual and prediced reurns. Abnormal reurns, e i,j,, for firm i, for year Published by Canadian Cener of Science and Educaion 129

5 j, on day are esimaed as he difference beween he acual reurn on day and he reurn expeced from he marke model. I hus represens he impac of firm specific even (dividend announcemens in his sudy) on shareholder wealh, ne of marke effecs. The abnormal reurns are calculaed as follows. e i, j, = Ri, j, α i, j β i, j Rm, j, By using he preceding equaion, daily abnormal reurns for each firm for each of he years 2005, 2006, 2007, and 2008 were compued over several even windows. Then, for any day wihin he even period he mean abnormal reurn (MAR) across sample members was calculaed as follows. MAR = N i = 1 e N i, j, where; e i,j, = abnormal reurn of securiy i for year j on day N = number of securiies Finally, cumulaive abnormal reurns over several holding periods from day 1 o day 2 were calculaed according o he following formula. = 2 1, 2 = 1 CAR MAR Under he null hypohesis ha dividend announcemens have no impac on corresponding sock prices, cumulaive abnormal reurns have an expeced value of zero. To es he hypohesis, parameric -ess are used. 4. Empirical Resuls As our aim was o invesigae he marke reacion o he change in dividends, he 184 announcemens by 46 companies were divided ino hree groups by comparing he dividend per share (DPS) for each observaion a ime o he previous dividend per share a ime -1. The groups were formed as follows. If DPS > DPS -1, he observaion is classified in he increasing dividends group If DPS < DPS -1, he observaion is classified in he decreasing dividends group If DPS = DPS -1, he observaion is classified in he consan dividends group As a resul, 88 announcemens were classified as increasing dividends, 73 as decreasing dividends and 23 as consan dividends. The resul of -es analysis on CAR values for hese groups is presened on Table 1. As can be seen, for he hree day even window surrounding he dividend announcemen, CAR(-1,1) value is posiive and saisically significan for he increasing dividends group, and negaive and saisically significan for he decreasing dividends group. For he consan dividends group, he CAR value is no significanly differen from zero. Therefore, he null hypohesis ha dividends have no impac on sock prices can be rejeced and our evidence is consisen wih he signaling hypohesis. However, such a resul was no obained for longer even windows, which resuled in saisically insignifican CAR values for all hree subsample of firms. Moreover, he pre-even period (-5,-1) shows significanly negaive CAR values. This means ha he fac ha an imporan announcemen will ake place is released o he public by he insiders before he acual announcemen dae. However, his only holds for decreasing dividends sample, suggesing ha bad news generae more reacion among insiders, inciing hem o disclose heir knowledge earlier. 5. Conclusion This sudy esed wheher dividend announcemens have an informaional conen for he Isanbul Sock Exchange. By applying he even sudy mehodology on a sample of 184 dividend announcemens made by 46 companies during he period 2005 o 2008, he average abnormal reurns and cumulaive abnormal reurns were calculaed. These calculaions were done for hree subsamples of securiies, namely hose wih increasing dividends, hose wih decreasing dividends and hose wih consan dividends. When a hree-day (1 day prior o he even day and 1 day afer he even day) even window was used, he resuls showed ha he marke reaced posiively o dividend increases, negaively o dividend decreases and did no reac when dividends were no changed. Therefore, he resuls were consisen wih he signaling hypohesis. 130

6 However, cumulaive abnormal reurns were insignifican when longer even windows were used. Also, he resuls showed pre-even informaion leakage for he decreasing dividends sample. Besides providing addiional evidence consisen wih he informaion conen hypohesis, hese resuls have imporan implicaions for companies. Specifically, firms should be careful when announcing changes in heir dividend policies since such announcemens have an impac on sock prices. References Aharony, J. & Swary, I. (1980). Quarerly Dividend and Earnings Announcemens and Sockholders Reurns: An Empirical Analysis. The Journal of Finance, 35 (1), Amihud, Y. & Li, K. (2006). The Declining Informaion Conen of Dividend Announcemens and he Effecs of Insiuional Holdings. Journal of Financial and Quaniaive Analysis, 41 (3), Asquih, P. & Mullins, D. W. (1983). The Impac of Iniiaing Dividend Paymens on Shareholders Wealh. The Journal of Business, 56 (1), Benarzi, S., Michaely, R. & Thaler, R. (1997). Do Changes in Dividends Signal he Fuure or he Pas?. The Journal of Finance, 52 (3), Bhaarcharya, S. (1979). Imperfec Informaion, Dividend Policy, and he Bird in Hand Fallacy. The Bell Journal of Economics, 10 (1), Chares, G. (1978). Dividend Informaion, Sock Reurns and Marke Efficiency-2. Journal of Financial Economics, 6 (2), Dyl, E. A. & Weigand, R. A. (1998). The Informaion Conen of Dividend Iniiaions: Addiional Evidence. Financial Managemen, 27 (3), Fama, E. F., Fisher, L., Jensen, M. C., & Roll, R. (1969). The Adjusmen of Sock Prices o New Informaion. Inernaional Economic Review, 10 (1), Fama, E.F. & French, K. R. (2001). Dissapearing Dividends: Changing Firm Characerisic of Lower Propensiy o Pay. Journal of Financial Economics, 60 (1), Garre, I. & Priesly, R. (2000). Dividend Behaviour and Dividend Signaling. Journal of Financial and Quaniaive Analysis, 35 (2), Grullon, G., Michaely, R,. Benarzi, S. & Thaler, R. (2005). Dividend Changes Do No Signal Changes in Fuure Profiabiliy. Journal of Business, 78 (5), Gunasekarage, A. & Power, D. M. (2002). The Pos-Announcemen Performance of Dividend-Changing Companies: The Dividend Signalling Hypohesis Revisied, Journal of Accouning and Finance, 42, Kalay A. & Loewensein, U. (1985). Predicable Evens and Excess Reurns: The Case of Dividend Announcemens. Journal of Financial Economics, 14, Kohari, S. & J. Warner. (2004) "Economerics of Even Sudies" Handbook of Corporae Finance, Vol. A,Ch.1. Linner, J. (1956). Disribuion of Incomes of Corporaions among Dividends, Reained Earnings and Taxes. American Economic Review, 46(2), MacKinlay, A. C. (1997). "Even Sudies in Economics and Finance." Journal of Economic Lieraure, 35(1), Michaely, R., Thaler, R. H. & Womack, K. L. (1995). Price Reacions o Dividend Iniiaions and Omissions:Overreacion or Drif?. The Journal of Finance, 50 (2), Miller, M. H., & Modigliani, F. (1961). Dividend Policy, Growh, and he Valuaion of Shares. The Journal of Business, 34 (4), Pei, R.R. (1972). Dividend Announcemens, Securiy Performance, and Capial Marke Efficiency. The Journal of Finance, 27 (5), Was, R. (1973). The Informaion Conen of Dividends. Journal of Business, 46 (4), Yılmaz, M. K. & Gunay, G. (2006). Dividend Policies and Price-Volume Reacions o Cash Dividends on he Sock Marke. Emerging Markes Finance & Trade, 42 (4), Published by Canadian Cener of Science and Educaion 131

7 Table 1. CAR Values for seleced even windows Dividend no change (N=23) Dividend increase (N=88) Dividend decrease (N=73) CAR (%) -sa CAR (%) -sa CAR (%) -sa CAR(-5,5) CAR(-3,3) CAR(-2,2) CAR(-1,1) * * CAR(-5,-1) *** CAR(-3,-1) CAR(-2,-1) CAR(1,2) CAR(1,3) CAR(1,5) *,**, and *** denoe significance a 10%, 5%, and 1% respecively. 132