Dividend smoothing and the long-run stability between dividends and earnings in Korea

Transcription

1 Korea Universiy Dividend smoohing and he long-run sabiliy beween dividends and earnings in Korea Jin-Ho Jeong Professor of Finance Division of Business Adminisraion Korea Universiy

2 I. Inroducion The signaling model of dividend -A firm's dividend policy is hypohesized o convey he informaion abou is fuure prospecs ha is privaely observed only by insiders (i.e., Miller and Rock (1985), John and Williams (1985), John and Nachman (1987), and ec.). Theory -John and Williams (1985) have demonsraed how dividends can ac as a credible signal of he fundamenal earnings prospecs of he firm. Pracice -For dividends o be a credible signal, insider should mainain a sysemaic relaionship beween expeced earnings and dividends. Empirical Evidence -Dividend Smoohing -Linner (1956) found ha managers smoohly adjus firms dividends wih respec o earnings and mainain a long-run arge payou raio. Dividend signaling hypoheses -signaling role of dividend Dividend smoohing credible signal.

3 Whereas dividend smoohing and dividend signaling are well-esablished empirical facs, he empirical evidences are based principally on informaion colleced in he US marke. The dividend policies of corporaions differ significanly across counries due o a variey of insiuional and financial marke differences( Aivazian (2003)). The objecive of his sudy is o invesigae wheher Korean firms follow dividend policies as in developed markes in which dividend smoohing and dividend signaling become sylized facs.

4 A common problem for sudying he dividend signaling hypohesis and he dividend smoohing hypohesis focus on he signaling model of dividend (e.g. dividend announcemen sudies) or he smoohing model of dividend separaely (e.g. parial adjusmen models), bu no boh Exising empirical ess canno help us o disinguish beween dividends are explici signals of fuure earnings (he dividend signaling hypohesis) dividends are smoohed wih respec o earnings o be a credible signal (he dividend smoohing hypohesis). There have been no empirical aemps o esimae and verify he dividend-earnings relaion reflecing boh hypoheses. Exising empirical ess canno really idenify he ypes of mechanisms for he dividend-earnings relaion (e.g. he join consideraions of signaling and dividend smoohing).

5 In his sudy A coinegraion model o es boh hypoheses in order o provide beer insigh ino he dividend and earning relaion. This allows us o invesigaeno only he informaional conen of dividend bu also he underlying facoro make dividend signal o be credible. Research implemenaion 1)Tes he informaional conen of dividend. Consruc a model o show he predicive relaions among dividends and fuure earnings a he firm level. We are paricularly ineresed in he issue of wheher he model can deec a presence of iner-emporal relaions beween dividends and earnings. 2) Tes he dividend smoohing hypohesis using a non-linear framework involving dividends and earnings sream. Tess of he ime-series implicaions of dividend models in a coinegraing framework offer a disincive approach o examine he abiliy of each hypohesis o explain corporae dividend behavior and hence, idenify he ypes of mechanisms for he dividend-earnings relaion.

6 The use of he coinegraion model Idenify he long-erm equilibrium relaionship beween dividends and earnings. Because heoreical analysis specifies he long-run equilibrium relaionship beween he earnings and dividend ime series in he presence of informaion asymmery, he coinegraion model employed in his sudy is regarded as a more direc es of he heoreical argumens in his area. Also a es is performed o examine he relaionship beween his derived dividend seing behavior of a firm and he informaion environmen usingboh mulivariae and univariae echniques.

7 II. Model Descripion of he Dividend Payou Paern Assume ha he dividend payou a ime can be described as follows. DPS = P * ( E [ EPS M, I ]) + 1 β (1) EPS + 1 = EPSZV + 1, Z I,and EPS M (2) where DPS is dividend per share, is earningper share, is a public EPS informaion se, I is an insider's informaion se a ime, Eis he expecaion Z V + 1 operaor, is a signal of he firm's +1 earnings observed a ime, is an innovaion erm wih mean 1, which is independen of boh publicand privae informaion a ime, and P can be inerpreed as a dividend payou raio when β = 1. M

8 Dividends are based on he earnings observable only o insiders condiioned on he rue qualiy of a firm. If a firm mainains a consan dividend payou raio, hen β is equal o one. β less han one indicaes ha dividends are a concave funcion of expeced fuure earnings condiioned on he insider's privae informaion hereby implying dividend smoohing. Subsiuing (2) ino (1) implies ha DPS = P * (E[EPS Z V M, I ]) + 1 β (3) Because EPS and Z are measurable wih respec o managemen's ime informaion which is he join of he public informaion signal ( Z ) E[EPS Z V + 1 M,I ] = EPS Z M ) ( E[V + 1 and heir privae M, I ]

9 Because he random innovaion erm is independen of M and I and has mean 0, E[V + 1 M, I ] = 1 Hence we can express (3) as DPS = P * (EPS Z ) β (4) The degree of smoohing can be measured by he curvaure of equaion (4). For a β Curve DPS = P ( EPS, Z ) he curvaure is given by ( 1 β) / EPS. β can be * esimaed by aking he naural log of equaion (4) o yield hefollowing esable Relaionship ln DPS = α + β ln EPS + ε (5) where ε = β1n(z ) and α = 1n (P). Equaion (5) will be used for finding he equilibrium relaionship beween dividends and earnings.

10 Idenificaion of Equilibrium Relaionship Beween Earnings and Dividends -Coinegraion If here exiss a long-erm equilibrium relaionship beween dividends and earnings, hen he deviaions from his relaionship, e in equaion (5) will follow a saionary process even hough dividends and earnings follow non-saionary process. The coinegraed variables ln DPS and 1nEPS exhibi a long erm equilibrium relaionship defined by ln DPS = α + βln EPS + ε and is he equilibrium error, which represens shor erm deviaions from he long-erm relaionship. β, coinegraing vecor is regarded as an underlying force o make ln DPS and o mainain a long erm equilibrium relaionship. ε 1nEPS Johansen (1988) coinegraion es is used Trace saisic denoed by LR r, maximum eigenvalue saisic denoed by λ max

11 III. DATA Sample Selecion and Daa Collecion We selec a sample of firms lised on he Korea Sock Exchange over he 26-year period from 1981 o The following sample selecion crieria were employed: 1.Firms had o have a leas 15 years of earnings and dividend daa during he period , as repored in he Korea Lised Companies Associaion daabase. 2.When esimaing Eq. (1), all firms wih non-posiive EPS or zero dividends were eliminaed from he sample, in order o preven he spurious resuls of dividend smoohing. 3.A furher screen excluded firms wih less han 15 observaions for each firm characerisic variable used in he regression. From a oal of 732 firms, 226 firms fulfilled hese screening crieria.

12 <Table 1> Sample Saisics Variable* Mean Sandard Deviaion Min Max DE SEPS SDPS EPS DPS LOGD LOGE * Dividend payou raio (DE) is measured by dividends per share divided by earnings per share. Earnings variabiliy (SEPS) is measured using he sandard deviaion of earnings per share overa minimum of 15 years prior o 2006 and a maximum of he weny-six years. Dividends variabiliy (SDPS) is measured using he sandard deviaion of dividends per share over a minim um of 15 years prior o 2006 and a maximum of he weny-six years. DPS is dividends per share and EPS is earnings per shar e. LOGD and LOGE sand for naural logs dividends and earnings, respecively.

13 Korean firms paid 468 KRW (or 0.40 USD) per share. The average dividend payou raio is 40.8%, which is lower han ha of US firms. The lower observed payou raio in Korea may be aribuable o he differen ax reamen of dividend income ax relaive o he capial gains ax. Korean firms appear o have less incenive opay dividends, as he unfavorable ax reamen of dividend income over capial gain is more serious han is he case in he US. In addiion, many invesors in Korea disregard dividends and consider sock price appreciaion as he principal componen of sock reurns. Korean invesor s aiude oward dividends may also conribue o lower dividend payou raios. Dividends variabiliy (SDPS) is while earnings variabiliy (SEPS) is Earnings variabiliy is approximaely 7 imes higher han dividends variabiliy, implying he presence of dividend smoohing. Average dividend payou raio from 1930 s o 2000s is 54.3% for S&P 500 companies.

14 IV. EMPIRICAL RESULT Coinegraion Tes Resul <Table 2> Summary Saisics for Equaion 5 Toal Sample (N=226) Coinegraing Firms (N=60) Noncoinegraing Firms (N=166) Variable* Mean Sandard Deviaion Min Max b R-squared * β is he coefficien of ln(eps) in equaion (5). I measures degree of dividend smoohing. R-squared is he explanaory powe r of he model. Coinegraion=1 if dividends and earnings are coinegraed

15 For coinegraion es -166 firms (73.5% of he oal sample) fail o rejec he null hypohesis of no coinegraion while 60 firms (26.5% of he oal sample) rejec he null concluding coinegraion. -Average values for β and R-squared are 0.67 and 58.5%, respecively. -The average magniude of he coefficien esimaor, β, indicaes ha on average, a firm mainains dividend smoohing as is dividend policy. -Average explanaory power of he model indicaes ha dividend changes are reasonably well represened by earnings.

16 Join Tes Resul for he Dividend Signaling and he Dividend Smoohing Hypoheses 1. Presence of coinegraion -The dividend signaling hypohesis suggess ha here will be along-erm equilibrium relaionship beween dividend changes and earning changes. Therefore, if here is a deviaion from he conemporaneous relaionship beween earning and dividends, here will be an underlying force o make hem o reurn o he equilibrium relaionship. 2. Magniude of coinegraing vecor -According o he dividend smoohing hypohesis, his underlyingforce will be a dividend smoohing. For coinegraing firms, he degree of dividend smoohing is expeced o be higher han ha of noncoinegraing firms. Table 3 shows he resuls.

17 <Table 3> Mean Differences of DDS Beween Coinegraing firms and Noncoinegraing Firms Coinegraing Firms (n=60) Noncoinegraing Firms (n=166) Variable Mean Sd.Dev. Mean Sd.Dev. DDS T-saisics for Mean Differences 1.9* (P-value=0.054) *DDS is he degree of dividend smoohing. I is measured by 1- ßin equaion (5).

18 Cross-secional properies of coinegraing firms and noncoinegraingfirms implied by he dividend signaling hypohesis. -The dividend signaling hypohesis suggess ha he higher degree of informaion asymmery is more likely o induce firms o mainain a long-erm equilibrium relaionship beween dividends and earnings. -To describe a firm s informaional environmens, we use lising years and firm sizeas proxies for he informaion environmen. <Table 4> Mean Differences of Informaive Variables Beween Coinegraing firms and Noncoinegraing Firms Coinegraing Firms Noncoinegraing Firms Variable Mean Sd.Dev. Mean Sd.Dev. T-saisics for Mean Differences SIZE (P-value=0.0008)*** HISTORY (P-value=0.0003)*** * Significan a he 10% level. **Significan a he 5% level. *** Significan a he 1% level. Firm size (SIZE) is esimaed by he naural logarihm of average oal asses during he sample period. Lising years (HISTORY) is measured by he number of years li sed during he sample period.

19 Logisic Regression Resul Based on he dividend signaling hypohesis, coinegraing firmsare hypohesized o face a greaer degree of informaion asymmery han noncoinegraing firms. We assign one o he coinegraing firms, and zero o noncoinegraing firms. SIZE and HISTORY have a significanly negaive relaionship wih he probabiliy of being coinegraing firms. The resul Table 5

20 <Table 5> Resul of Logisic Regression by Using Informaive Variables as Independen Variables Dependen Variable ; Transformed Probabiliy of Being Coinegraing firm Explanaory Variable Coefficien Chi-square Inercep *** SIZE *** HITORY *** Chi-square = 20.90*** * Significan a he 10% level. ** Significan a he 5% level. *** Significan a he 1% level. Firm size (SIZE) is esimaed by he naural logarihm of average oal asses during he sample period. Lising years (HISTORY) is measured by he number of years lised during he sample period. The resul implies ha younger and smaller firms have a endency o have a coinegraing relaionship beween earnings and dividends.

21 Robusness check oher consideraions <Table 6> Resul of Logisic Regression by including conrol variables Dependen Variable ; Transformed Probabiliy of Being coinegraing firm Explanaory Esimae Error Chi-Square Pr Variable Inercep *** SIZE ** HISTORY *** SLACK GROWTH LEV SEPS *** LARGE Chi-square = 31.87*** * Significan a he 10% level. ** Significan a he 5% level. *** Significan a he 1% level. Firm size (SIZE) is esimaed by he naural logarihm of oal asses, growh rae (GROWTH) is he average growh rae of size over a minimum of 15 years prior o 2006 and a maximum of he weny six years, financial slack (SLACK) is he raio of accumulaed reaine d earnings o oal asses, and earnings variabiliy (SEPS) is measured using he sandard deviaion of earnings per share over a minim um of 15 years prior o 2006 and a maximum of he weny-six years. Lising years (HISTORY) is measured by he number of years lis ed during he sample period. The percenage of he sock held by he larges shareholders (LARGE) is used as a proxy for he concenr aion of conrolling shareholders. Leverage raio (LEV) is measure by he mean leverage raio (oal liabiliies divided by oal asses) f or a minimum of 15 years prior o 2006.

22 V. CONCLUSIONS There have been no empirical aemps o esimae and verify he dividend-earnings relaion reflecing boh he signaling and dividend smoohing hypoheses. This sudy proposed a coinegraion model o es boh hypoheses in an inegraed framework in order o provide beer insigh ino he dividend and earning relaion. This sudy pursued he following wo main objecives: 1. wha is he characerizaion of he opimal dividend payou paerns on he individual firm basis 2. wha is he effec of informaion on he firm's dividend payou policy. The resuls of -es and logisic regression - The presence of coinegraion is posiively relaed o he degree of informaion asymmery, a resul consisen wih he dividend signaling hypohesis - Dividend smoohing is idenified as an underlying force o makedividends and earnings coinegraed.

23 Suggesions for Fuure Research Tes of Residual Theory The residual heory holds ha dividend policy is deermined by he need for funds o finance invesmen. Dividends are a residual ha reflecs he amoun lef over from earnings afer invesmen projecs are financed by equiy capial. The following coinegraing regression migh be useful o see wheher here is an equilibrium relaionship beween equiy capial and dividends. DIVIDENDS = β 0 + β1earnings + β 2EQUITY CAPITAL Since we expec a perfec negaive relaionship beween dividends and equiy capial, β 2 he esimaor of is expeced o be negaive 1.