Share Repurchase Behavior of Japanese Banks

Transcription

1

2 Share Repurchase Behavor of Japanese Banks Takash Hatakeda 1 Graduate School of Busness Admnstraton, Kobe Unversty, -1 Rokkoda, Nada, Kobe, Hyogo , Japan May, 006 Abstract Accordng to Artcle 10 of the Commercal Law, a bank (or a frm) wth an ntenton of repurchasng shares has to make two decsons at the board of drectors. The one s the decson about whether a bank has a tendency of repurchasng shares n the future or not and then about how many shares she repurchase f she does so. The other s the decson about how many shares she repurchases shares actually n the authorzaton perod. In our paper we show the emprcal evdence that non-performng loans and (or) the captal to asset rato have a crucal nfluence on two decsons n the Japanese bankng ndustry. JEL classfcaton: G14; G1; G3; Keywords: Bank; Stock repurchase; non-performng loans; the captal to asset rato 1 Tel: E-mal address: hatakeda@kobe.u-ac.jp

3 1. Introducton Though Japanese commercal law had banned domestc frms from repurchasng ther own shares untl October 1994, the revson of the commercal law made t possble for them to buy back and cancel ther outstandng shares n order to dstrbute earnngs to shareholders. As, however, there were severe obstacles whch prevent them from utlzng share repurchases as a payout polcy, the commercal law was more revsed at frequent ntervals and the Specal Case Law of share repurchases was also ntroduced. As a result, share repurchases have becomng an mportant method of the payout polcy for corporatons n Japan smlar to U.S. 3 Snce October n 001, frms wth an ntenton of repurchasng shares have to undergo the followng process at the board of drectors and the shareholders meetng. In the frst place, a frm has to ntate a resoluton on share repurchases at the begnnng of the fscal year. The target yen amount of a repurchase program, the target number of shares, the knd of shares, and so on are stpulated n the ntated resoluton, whch must be approved n the shareholder meetng. Secondly, she can be free to decde to repurchase ts own shares on the ground of the approved resoluton. There, thus, are two decsons about share repurchases; a plannng decson and an executon decson. It, In Appendx A, we llustrate the regulatory and nsttutonal background of share repurchases n Japan. 3 See, for example, Vermaelen (1981), Barclay and Smth (1988), Bagwell and Shoven (1989), Wansley and Sarkar (1989), Comment and Jarrell (1991), Ikenberry, Lakonshok and Vermaelen (1995), Cudd, Duggal and Sarkar (1996) Lu and Zebart (1997), Grnblatt and Ttman (1998), and Grullon and Mchaely (00). For example, Grullon and Mchaely (00) report that, over the past 0 years, stock repurchase expendtures grew at a much hgher rate than cash dvdends. In Japan, 48% of all frms whch have been lsted on the Frst Secton of the Tokyo Stock Exchange made share repurchase announcements n the perod from October, 001 to December, 00 (See Hatakeda (005)).

4 therefore, s nterestng for us to examne what factors affect two decsons on share repurchases. Ths revson n 001 was also drastc n a sense that t has made t possble for frm s managers not only to buy back ther own shares wthout declarng the purposes but also to hold repurchased shares wthout cancelng or resellng them at once. It seems that ths revson enhances the avalablty of repurchasng shares. Indeed, accordng to aggregate data from IN Informaton System Ltd, the number of the repurchase announcements ncreased sgnfcantly from 485 over four years to 1361 n two years of In addton, the total number of shares whch have been nscrolled n the repurchase programs n 001 and 00 ncreases.5 tmes as large as the total amount over the years of The purpose of our paper s to manfest what factors have a crucal nfluence on the plannng decson and the executon decson about share repurchases n Japanese bankng ndustry. As we have mentoned before, t s thought that the revson of the Commercal Low n 001 have a dfferent mpact on a share repurchase behavor from before. We, therefore, focus on the repurchase behavor snce October 001 n order to solate the structural effect due to the revson of the Commercal Law. The dstnctve feature of our paper s to shed a lght on the repurchase behavor n the Japanese bankng ndustry. The ncreased promnence of share repurchases n the bankng ndustry mrrors the growth of share repurchases by non-fnancal frms, whch has been documented by a number of studes (see, for example, Zang (00), Hatakeda and Isagawa (004) n Japan), so one queston that arses s whether analyss of share repurchases by fnancal frms n partcular, banks would offer any further nsghts or not. At the same tme, the crcumstance n 3

5 whch banks operate has changed drastcally n ths decade. Untl the early 1990s the Japanese fnancal system was often mentoned as havng typcal bank-based system n the contrast wth the market-based systems n the U.S. The Japanese government was also nstrumental n the development of the man bank system. Japanese banks had enjoyed monopolstc rent n a bank-based fnancal system untl the early of 1990s (.e. man bank system, the convoy system between banks, cross share holdngs). Banks, however, have experenced the 1990s economc stagnaton and the drastc change nto a market-based fnancal system. 4 Bank s rent has been extngushed by a huge of non-performng loans and by cutthroat competton among fnancal ntermedatons. Furthermore, the decrease n Japanese bank s latent captal gans, whch s called as Fukum Ek, has caused a lot of damage to bank s equtes for ths decade, so that banks suffered from the low captal asset rato. 56 Accordngly, t s nterestng to examne whether or not non-performng loans and bank s equtes affect her decsons regardng the target for share repurchases and ts executon. The remander of our paper s organzed as follows. In secton, we dscuss the data set used n our study and present the descrptve statstcs. In secton 3 and 4, we 4 Large frms that had ever had a close relatonshp wth banks changed from bank fnancng to captal-market fnancng. It s supposed that A great declne n bank s share prces makes banks reduce the share of the bank-shareholdng and the number of board members accepted from banks. Accordng to Hosh and Kashyap (001), n the perod between 1975 and 1997, the average bank-borrowng to sales rato declned from 35 % to %, the average rato of frms for whch a bank s the largest equty holder declned from 13 % to 9 %, and the average rato of frms wth any bankers on ther Boards declned from 55 % to 51 %. As a result, Japanese banks could not help but ncrease the share of assets allocated to lendng to small and medum frms. 5 In and 001 the Japanese economy has experenced the most serous recessons. Standard & Poor s Ratng Servce reports the average ratng have deterorated by BBB snce 1998, compared to AA mnus n Latent captal gans are defned as the dfference between the low book value and the market value of shares whch have been held by banks. The Japanese banks tradtonally have held a large amount of stocks of other companes as a part of long-term busness relatonshp, keretsu. 4

6 estmate our models that relate varous factors to bank s decsons regardng the target for repurchase shares and regardng ts executon, and then dscuss our results obtaned, respectvely. Fnally, n secton 5, we contan summary and conclusons.. Data and Unvarate Analyss of Repurchases Usng the Nkke Needs Fnancal Quest, we extract all the commercal banks that appear on the fles n the fscal years of Our sample conssts of 95 banks that were traded n the 1st and nd Secton of the Tokyo, Osaka, Fukuoka, Nagoya Stock Exchange, and over the counter market (OTC) n the perods from Aprl 1, 000 to March 31, 003. They are composed of Cty banks (5 banks) and Regonal banks (90 banks). 7 We delete commercal banks wth mssng or nconsstent data and those that experenced large mergers and acqustons. In Japan, the removal of a ban on fnancal holdng companes n March 1998 led to large mergers and acqustons n Cty banks and some of regonal banks by March 00. Consequently, our sample ultmately conssts of 85 Regonal banks wthout a fnancal holdng company structure. Table 1 reports summary statstcs n the fscal years of for the full sample and for the two subsamples of Japanese banks: Total, Repurchase Group[RG], 7 Cty banks have now adopted a fnancal holdng company structure and have been extendng ther busness to the varous areas. Although they prmarly have played a role of the supplers of short-term funds to large corporatons, they have recently begun to focus on developng the longer-term end of ther busness. Addtonally, they have been forced to cultvate clents from amongst the smaller corporate fry and from the personal sector. The top four cty banks, by assets sze, are some of the largest banks n the world. Regonal banks are smaller n scale than Cty banks, and usually confne ther operatons to the prncpal ctes of the prefectures n whch ther head offces are located. Accordngly, ther local tes are strong, wth the bulk of ther lendng gong to small and medum-szed companes n the localty. 5

7 and Non-Repurchase Group[NGR]. Here, RG conssts of 45 banks wth an ntenton of repurchasng shares. On the other hand, NRG conssts of 40 banks wth no ntenton. 8 On average, banks wth a large asset sze ( SIZE ), current low-proftablty ( ROA ) and stable nvestment opportuntes ( M / B ) tend to repurchase ther own shares. Addtonally, banks wth a low nonperformng loan rato ( NPL L ) 9, a low loan rato for the small and medum frms ( SMALL ), and a hgh captal-asset rato ( ECR 10 ) 8 Accordng to Artcle 10 of the Commercal Law, a bank wth an ntenton of repurchasng ts own shares has to pass through the followng steps. Frst of all, the board of drectors decdes to ntate a repurchase program at the begnnng of the fscal year. And then the board sends a notce to the stock change and makes a publc announcement of the repurchase plan through mass meda (TV, newspapers, rado, and so on) wthout delay. Secondly, the proposed resoluton to repurchase shares has to be approved at shareholder meetng. Untl the next year s shareholders meetng, the bank can be free to carry out share repurchases based on the approved resoluton. 9 There are three measures of non-performng loans (NPL/L) n Japan. The frst measure s rsk management loans, whch s publshed by each bank on ts fnancal statement. The second s self-classfed loans, whch s based on the classfcaton of loans that the Fnancal Supervsory Agency uses n bank examnatons. The thrd s classfed loans by the Fnancal Reconstructon Commsson. Though rsk management loans have been made avalable begnnng wth the year ended March 1993, ts defnton has changed by March Although the defnton by the second measure s substantally broader than that of the frst and thrd measures, a majorty of non-performng loans defned by the second measure are consdered recoverable. The thrd measure s slghtly broader than that of rsk management loans (For further detals of three defntons of non-performng loans, see Hosh and Kashyap (1999, 001)). We use rsk management loans whch are publshed by each bank on ts fnancal statement as non-performng loans. Rsk management loans consst of loans to faled enterprses, loans on whch payments are suspended for more than three months, and loans wth relaxed condtons (restructured). As of March 00, rsk management loans totaled just over 4 trllon, or 7.9% of Real GDP. 10 Unlke other ndustres, n order to support safe and sound bankng system, the bankng ndustry s requred to keep a flat mnmum percent of captal aganst rsk-weghted assets, whch are the sum of the total weghted assets where weghts are set larger for the rsker assets. (See Hall (1993)). In Japan, there are two types of banks: nternatonal and domestc banks. They are held to dfferent standards respectvely. The Basel Accord requres nternatonally actve banks to mantan a captal to rsk-weghted asset rato of at least 8%. On the other hand, Japanese regulators allowed banks wth purely domestc busness the opton of mantanng the captal rato of at least 4%. We defned ECR as the rsk-based captal rato reported n bank s fnancal statement less the nternatonal standard (8%) or the domestc standard (4%). 6

8 do so. Fnally the percent of equty held by 10 largest shareholders ( SHARE ) n RG s smaller than that n NRG. However, ts volatlty n NRG s relatvely larger than that n RG. [ Yen] Table 1 also reports two repurchase target ratos n RG; TAR [Number] and TAR 11, and two executon ratos; [Number] EXR and EXR [ Yen] 1 n the fscal years of 00. Banks n RG set up the mean target rato to 4% and execute only 5% of the target. Both ratos are lower than those n frms excludng from fnancal ntermedares. For example, Hatakeda (005) reported that the mean target rato s 6 % and that the mean executon rato s 40%. 13 It s consdered that ths low executon rato s led to by severe envronments whch surround Japanese banks, such as the deteroraton of share prces and land prces, a large number of accumulated non-performng loans, and the low captal asset rato due to the economc stagnaton, and then dsntermedaton, whch large frms n Japan tend to fnance funds by ssung equtes rather than bank loans. Before dscussng what factors affect bank s decsons regardng share repurchases, we should examne the possblty that banks make use of cash dvdends as substtutes for share repurchases n order to return banks own profts nto ther 11 TAR[Number] represents the rato of the number of the targeted shares to the total number of outstandng shares. TAR[Yen]represents the rato of the yen value of the targeted shares to the end-of-perod value of the outstandng shares. 1 EXR[number] s expressed as a rato of the number of shares executed to repurchase aganst the number of shares targeted. EXR[Yen] s expressed as a rato of the market value of shares executed to repurchase aganst the yen value of the target shares. 13 Stephens and Wesbach (1998) reported the mean target rato was 7% and that the mean executon rato was 54% for open-market repurchase programs n US Market for the perod from 1981 to In partcular, the executon rato n Japan s lower than those n U.S. Thus, although Japanese frms made a large number of repurchase announcements, they could repurchase a few shares actually. It seems that ths low executon rato s due to economc stagnaton n the perods from 001 to

9 shareholders. If cash dvdends are substtutes for share repurchases n the Japanese bankng ndustry, banks whch pay out less cash dvdends or omt cash dvdends mght attempt to set up the hgh target for share repurchases and execute to repurchase more shares. 14 Table s descrptve statstcs concernng cash dvdends and share repurchases across two groups n the fscal years of From Panel (1) and Panel (), we fnd no evdence that share repurchases are substtutes for cash dvdends. 15 Although dvdend-payng banks both n RG and n NRG make the same payments per a unt share, banks whch pay cash dvdends tend to repurchase ther own shares n order to dstrbute more proft to shareholders. In Panel (3), furthermore, we can obtan the evdence that there exsts a sgnfcant and postve correlaton between actually repurchased amounts and dvdend payments. In sum, our evdence suggests that there s no substtuton between share repurchases and dvdends. It, f anythng, seems that banks make use of share repurchases as a complementary method of payng out profts avalable for dstrbuton. 3. Decson Regardng the Repurchase Program Accordng to Artcle 10 of the Commercal Law, a bank (or a frm) wth an ntenton of repurchasng shares has to make two decsons at the board of drectors. The one s the decson about whether a bank has a tendency of repurchasng shares n 14 Emprcal results for ths substtuton hypothess are mxed. For example, Bagwell and Shoven (1989), Grullon and Mchaely (00) reported the evdence that dvdends and repurchases are substtute n U.S. On the other hand, Jagannathan et al. (000), DeAngelo et al (000), Fama and French (001) reported the opposte evdence that there was no substtuton n U.S. 15 There s a statstcally sgnfcant dfference n the mean (medan) of the ratos across two groups 8

10 the future or not and then about how many shares she repurchase f she does so. The other s the decson about how many shares she repurchases shares actually n the authorzaton perod. In ths secton we focus on the former decson. And then we dscuss the latter decson n secton 4. We estmate several knds of regresson models whch relate bank s ndvdual-specfc characterstcs to the frst decson on ts repurchase plan. We construct the repurchase target ratos; [Number] TAR and [ Yen] TAR as a observed dependent varable y. TAR [Number] represents the rato of the number of the targeted shares to the total number of outstandng shares as of March 31, 00. TAR [ Yen] represents the rato of the yen value of the targeted shares to the market value of the outstandng shares as of March 31, 00. These are censored varables of whch the value are larger than zero f the decson regardng the share repurchase program s made by the board of drectors, zero otherwse. Therefore, our model s y = β ' x + u, (1) * f y * > 0, y = [Number] TAR or [ Yen] TAR f y * 0, y = 0. where * y denotes the latent varable, x denotes a vector of explanatory varables lkely to affect the decson regardng the target for repurchase shares, and u denotes the error term whch follows a normal dstrbuton wth mean zero and varance σ. 9

11 Followng to a large number of lterature about frm s payout polcy 16, we employ the followng varables as explanatory varables; a constant term, the log of total assets ( ln ( SIZE) ), the current proftablty ( ROA ), and the market-to-book rato ( M / B ). In our paper, M / B s served as a proxy for growth opportuntes. Unlke other types of ndustres, the bank ndustry s subject to the nfluence of some regulatons and oblgatons whch requre them to carry out sound management. In order to examne these nfluences, we append two varables to our model. The one s the excess captal to asset rato ( ECR ), whch s calculated as the rsk-based captal rato reported n bank s fnancal statement less the requred rato whch the bank has to keep (4% or 8%). 17 The other s the rato of non-performng loans to total loans and blls dscounted ( NPL / L ). Our predcton s that banks wth low ECR and hgh NPL / L can not help resgnng to pay back shares or settng the lower target rato even f they have an ntenton wth repurchases. All of explanatory varables whch we utlze are the end-of-prevous-perod values, snce the board of drectors makes a decson on the repurchase plan at the begnnng of the fscal year. Therefore, all the explanatory varables are pre-determned. We estmate the parameters β and σ by the maxmum lkelhood (ML) method See, for example, Stephens and Wesbach(1998), Fama and French (001), DeAngelo, et al.(006). 17 See footnote The lkelhood functon for ths model (1) s gven by L 1 σ y β ' x σ ( ) β, σ y, = x * φ where φ () and () y = y y = 0 β ' x Φ σ. Φ are the densty functon and the dstrbuton functon of the standard normal, respectvely. We estmate the parameters β and σ by maxmzng the log-lkelhood functon. For detals of the censored regresson model, see Wooldrdge (00). 19 We, of course, nclude the constant term n estmatng our models. 10

12 In Table 3 we present estmaton results for the models n whch ether TAR [Number] or TAR [ Yen] s used as the dependent varable. In Panel (1) and Panel (3), we present the results for our benchmark model, whch the explanatory varables are a constant term, ln ( SIZE), ROA, and M / B. From our result that the coeffcents on ln ( SIZE) and M / B s sgnfcant at the 1% level, we fnd the evdence that small banks wth a low nvestment (lendng) opportunty are lkely to set up the hgher target for repurchasng ther own shares. In Panel () and Panel (4) we present the results for the augmented model wth ECR and NPL / L as addtonal explanatory varables. We fnd that the coeffcent on NPL / L s negatve and sgnfcant n both columns at the 1% level, although the coeffcent on ECR s postve but nsgnfcant n both columns. Our result suggests that an ncrease n non-performng loans reduces bank s avalable funds for repurchases because she has to ncrease the provsons for possble loans losses, so that banks wth large nonperformng loans can not help resgnng to pay back shares or settng the lower target for repurchase. Snce the early 1990s, banks have experenced a decrease n cross-share holdngs, wth a process of the deregulaton of Japanese captal market. 0 So, some banks may attempt to make use of share repurchases n order to mantan corporate governance desrable for them. To dscuss ths possblty, we re-estmate the model to whch we adds the change of the rato of outstandng shares owned by the largest shareholders ( Δ SHARE ) as an explanatory varable. Table 4 provdes the results for TAR [Number] and for TAR [Yen] as a dependent varable, respectvely. The coeffcents on ( SIZE) ln, ROA, M / B, ECR, and 0 See Okabe (00). 11

13 NPL / L are almost smlar to the pervous results. What should be noted s that the coeffcent on Δ SHARE s nsgnfcant n the case of both TAR [Number] and TAR [ Yen]. These results ndcate that, even f banks experence the declne of outstandng shares owned by large shareholders, they do not attempt to set up the hgh target for share repurchases n order to reman the captal structure unchanged. 4. Decson Regardng Repurchase Executon Banks whch have set the repurchase program do not necessarly have to execute repurchases untl the set target. Indeed, some banks execute no repurchases at all. In ths secton, we dscuss what factors affect a bank s decson regardng the executon of share repurchases. The estmaton procedure s almost smlar to that of secton 3, but s dfferent n three respects. The frst s that we use the executon rato n place of the target rato as a dependent varable. We utlze the fnancal statements n 003 to calculate two executon ratos; [Number] EXR and EXR [ Yen], whch represent the degree of actual share repurchases. The former s defned as the number of the repurchased shares and the latter s ts yen value. 1 These varables are also censored varables, whose the value are lager than zero f an executon decson s made, zero otherwse. Secondly, when a bank repurchases ts own shares actually, the board of 1 The former s expressed as a rato of the number of shares executed to repurchase aganst the number of shares targeted. The latter s expressed as a rato of the market value of shares executed to repurchase aganst the yen value of the target shares. In our sample, max value of the executon rato s less than 100% rrespectve of the executon rato. Addtonally, 19 of 40 banks whch have set the repurchase program dd not execute share repurchases after all. 1

14 drectors s lkely to carry out tmely wth watchng the trend of the share prce performance. To allow for ths factor, we use the cumulatve abnormal return from the day after the shareholder s meetng to the day before the executon ( LCAR 3 ) n place of the proxy for the proftablty to our model. 4 Some banks conduct the executon of repurchases n twce or more tmes. In calculatng LCAR, we focus only on the frst tme of the repurchase executon n order to avod the hstory effect. The thrd s that we use the explanatory varables are based on a bank s fnancal statements n 003, because banks carry share repurchases nto executon dependng on the fundamental and fnancal stuatons that they face at present. Therefore, n order to allow for the endogenety of the explanatory varables M / B, ECR, ROA, and Δ SHARE, we estmate the Tobt model wth endogenous varables 5 by the maxmum lkelhood estmaton. In estmatng our model, we utlze the constant term, ( SIZE) ln, LCAR and the one-lagged and two-lagged values of M / B, ECR, and SHARE as nstrumental varables. In Table 5 we present our estmaton results for [Number] EXR and EXR [ Yen] as a dependent varable. Although the coeffcent of ln ( SIZE) and M / B s not sgnfcant n both models, the coeffcent of LCAR s negatve and sgnfcant at the 1% level n both models. These results ndcate that a bank attempt to repurchase her shares actually, when her share prce declnes. The coeffcents of ECR are postve and sgnfcant at the 5% level for the 3 If a bank has not executed repurchase ts shares n the end, LCAR s defned as the cumulatve abnormal return from the day after the shareholder s meetng to the day before the next shareholder s meetng. 4 US s current dsclosure standards, as regulated by the FASB and the Securtes and Exchange Commsson (SEC), mpose an oblgaton on frms to dsclose the number of shares repurchased at quarter-end. On the other hand, n Japan Securtes and Exchange Survellance Commsson (SESC) requred frms to dsclose repurchase detal. In partcular, we can know the tradng dates. 5 See Appendx B. 13

15 case of dependent varable EXR [Number] and at the 10% level for the case of dependent varable EXR [ Yen]. These results ndcate that banks have a postve ncentve to make a decson to execute share repurchases when they have the suffcent volume of ther own captal. Furthermore, NPL / L are also postve and sgnfcant at the 5% level n both models. Thus, the volume of non-performng loans has large nfluence on the repurchase executon as well as ts plan. In Table 6 we present our estmaton results n order to examne the effect of the share rato by the largest shareholders ( Δ SHARE ) on the decson regardng the repurchase executon. The coeffcents of Δ SHARE n both models are nsgnfcant, suggestng that the declne of outstandng shares owned by the largest shareholders have less effect on the executon of share repurchases. It seems that banks do not make use of share repurchases as countermeasures aganst the cancellaton of the cross-shareholdngs. 5. Conclusons In our paper, we examned the factors whch affect a bank s decson regardng the repurchasng program and regardng the executon of repurchases. Accordng to our results, the degree of non-performng loans has a negatve effect on both decsons. Furthermore, the degree of the captal asset rato has a postve effect on the decson regardng the executon of repurchases. As our addtonal fndng, we examne the possblty that, when banks experenced the declne of cross-shareholdngs, they tend to set up the hgh target for share repurchases and execute repurchases actually n order to mantan the optmal 14

16 captal structure. However, we could not fnd the evdence that ths possblty s supported. Fnally, we should notce that our emprcal results may reflect the repurchase behavor of banks wthout a fnancal holdng company structure. Large mergers and acqustons such as a fnancal holdng company structure may have an effect on the bank s share repurchase behavor. These ssues are topcs of future research. Appendx A: Stock Repurchases of Japanese Frms Pror to 1994, Japanese frms had been n prncple prohbted to buy back ther outstandng shares. The revson of the Commercal Law (Artcle 1-) n October, 1994 has enabled frms to buy back ther outstandng shares on the bass of the resoluton whch had to be approved at a regular shareholder s meetng. Ths resoluton has been effectve untl the next shareholder s meetng. The funds used for repurchases have to come from the proft avalable for dvdends. However, under Japanese accountng rules of those days, a presumed dvdend tax was mposed n buyng back ther standng shares. In fact, share repurchases were effectvely nonexstent untl November, 1995, when the presumed dvdends for share repurchases have been frozen. In addton to Artcle 1- of the Commercal Law, the Japanese government enforced Artcle 3-1 of the Specal Case Law n June, 1997, Artcle 3- n March, 1998, and Artcle 3-3 n March, Under Artcle 3-1, the board of drectors was allowed to execute share repurchases wthout approval at a regular shareholders meetng once the agreement on repurchases was provded n the artcles. Artcle 3- and 3-3 allowed frms to use captal reserves and latent captal gans on lands for busness to repurchase shares. In contrast to Law Artcle 1- of the Commercal Law, Artcle 3 of the Specal Case Law had the advantage that the board of drectors was allowed to elastcally execute share repurchases. It, however, had the dsadvantage that the funds used for repurchases and the number of repurchased shares were restrcted. A large number of frms whch executed share repurchases were based on Artcle 3 of the Specal Case for the perod from 1999 to 001. Furthermore, the Commercal Law revsed n October, 001, so that the treasury shares were ntroduced and the Specal Case Law was dropped. Not only the 15

17 001 s revson of the Commercal Law had the advantage of the Specal Case Law, under whch frms elastcally execute share repurchases, but also t enabled them to hold shares repurchased for the perod. Ths mples that frms do not have to dspose repurchased shares mmedately, ndcatng that they also have the opportunty to wat for ther shares to rse before resellng ther shares. Appendx B: The estmaton of the Tobt model wth endogenous varables In estmatng the Tobt model wth endogenous varables, we formulate the followng smultaneous model: * 1 y = γ ' x + δ ' x + u 1 1 x = π ' x + π ' z + v where * y denotes the latent varable, x 1 s a vector of predetermned varables, s a vector of endogenous varables, and z s a vector of exogenous (nstrumental) varables. ( u, v ) s normally dstrbuted wth mean zero and varance, x σ ρ Σ =. ρ τ 1 In our model, ln SIZE and LCAR, x denotes a vector of M / B, ECR, NPL / L, and Δ SHARE, and z denotes a vector of the frst and second lagged varables of M / B, ECR, NPL / L, and SHARE. The observed dependent varable EXR [Number] x denotes a vector of one, ( ) y = or EXR [ Yen] otherwse. The lkelhood functon for ths model s gven by * > y = s the value f y 0, y = 0, L 1 ( γ δ, π, π, σ, τ, ρ y, x ) = y φ y = 0 1 π ' x γ ' x Φ 1 1 π ' z, * y = y τ δ ' x η θv 1 y φ τ 1 γ ' x 1 δ ' x η θv where η ρ = σ and τ ρ 1 θ =. We omt the estmated results of coeffcents π ' and τ 16

18 π ' from Table 5 and 6 for want of space. [ ] References [1] Bagwell, L., and J. Shoven, 1989, Cash Dstrbutons to Shareholders, Journal of Economc Perspectves 3, No.3, [] Barclay, M., and C. W. Smth, 1988, Corporate Payout Polcy: Cash Dvdends versus Open Market Repurchases, Journal of Fnancal Economcs, [3] Brown, S. J., and J. B. Warner, 1985, Usng daly stock returns: The case of event studes, Journal of Fnance 14, [4] Comment, R., and G. Jarrell, 1991, The Relatve Sgnalng Power of Dutch Aucton and Fxed Prce Tender Offers and Open Market Share Repurchases, Journal of Fnance 46, [5] Cudd, M., Duggal, R., and S. Sarkar, 1996, Share Repurchase Motves and Stock Market Reacton, Quarterly Journal of Busness and Economcs 35, [6] DeAngelo, H., L., DeAngelo, and D., Shnner, 000, Specal dvdends and the evoluton of dvdend sgnalng, Journal of Fnancal Economcs, 57, [7] DeAngelo, H., L., DeAngelo, and R. M., Stulz, 006, Dvdend polcy and the earned/contrbuted captal mx: a test of the lfe-cycle theory, Journal of Fnancal Economcs, 74, forthcomng. [8] Fama, E., and K., French, 001, Dsappearng dvdends: changng frm characterstcs or lower propensty to pay? Journal of Fnancal Economcs, 60, [9] Grnblatt, M., and S. Ttman., 1998, Fnancal Market and Corporate Strategy, McGraw Hll. [10] Grullon, G., and R. Mchaely, 00, Dvdend, Share Repurchases, and the Substtuton Hypothess, Journal of Fnance [11] Hall, B. J., 1993, How Has the Basle Accord Affected Bank Portfolos, Journal of the Japanese and Internatonal Economcs 7, [1] Hosh, T., and A. Kashyap., 1999, The Japanese Bankng Crss: Where dd t Come From and How wll t End?, n B. S., Bernanke and J. Rotemberg ed., NBER Macroeconomcs Annual [13] Hosh, T., and A. Kashyap., 001, Corporate Fnancng and Governance n Japan, The Road to the Future, MIT Press. [14] Hatakeda, T., and N. Isagawa, 004, Stock Prce Behavor Surroundng Stock 17

19 Repurchase Announcements: Evdence from Japan, Pacfc-Basn Fnance Journal 1, [15] Hatakeda, T., 005, Stock Prce Behavor Surroundng Stock Repurchase Announcements: Evdence from Japan snce October, 001, Workng Paper (In Japanese). [16] Ikenberry, D., Lakonshok, J., and T. Vermaelen, 1995, Market Underreacton to Open Market Share Repurchases, Journal of Fnancal Economcs 39, [17] Jagannathan, M., C., Stephens, and M., Wesbach, 000, Fnancal flexblty and the choce between dvdends and stock repurchases, Journal of Fnancal Economcs, 57, [18] Lu, C-S., and D. Zebart, 1997, Stock Returns and Open-Market Stock Repurchase Announcements, Fnancal Revew 3, [19] Okabe, M., 00, Kavbusk Mocha to Nhongata Keza Ssutemu (n Japanese), Keo Dagaku Syuppanka. [0] Stephens, C., and M. Wesbach, 1998, Actual Share Reacqustons n Open-Market Repurchase Programs, Journal of Fnance 6, 3-7. [1] Vermaelen, T., 1981, Common Stock Repurchases and Market Sgnalng: An Emprcal Study, Journal of Fnancal Economcs 9, [] Wansley, J., W. Lane, and S. Sarkar, 1989, Managements Vew on Share Repurchase and Tender Offer Premums, Fnancal Management 18 (autumn), [3] Wooldrdge, J. M., 00, ECONOMETRIC ANALYSIS OF CROSS SECTION AND PANEL DATA, MIT Press. [4] Zhang, H., 00, Share Repurchases under the Commercal Low 1- n Japan: Market Reacton and Actual Implementaton, Pacfc-Basn Fnance Journal 10,

20 Table 1. Summary of Statstcs Total Repurchase Group [RG] Non-Repurchase Group [NRG] Varables Standard Standard Standard Mean Medan Mean Medan Mean Medan Devatons Devatons Devatons SIZE (Bllon Yen) ROA (%) M/B (%) NPL/L (%) SMALL (%) ECR (%) SHARE (%) TAR [Number](%) TAR [Yen] (%) EXR [Number](%) EXR [Yen] (%) Number of Observatons Notes: SIZE s defned as the sum of the book value of debt and the market value of equty. Here, the market value of equty s defned as the product of the share prce and the number of outstandng shares. The share prce s the end of the fscal year. If the prce s not avalable for the last day of the fscal year, the next earlest avalable prce s used. M/B s defned as the percentage of (the sum of the book value of debt and the market value of equty)/(the sum of the book value of debt and the book value of equty). NPL s defned as the percentage of rsk management loans (the sum of loans to bankrupt borrowers, non-accrual delnquent loans, loans past due for 3 months or more, and restructured loans) / total loans and blls dscounted. ECR s defned as the rsk-based captal to asset rato mnus the nternatonal standard (8%) or the domestc standards (4%). SHARE s defned as the percentage of outstandng shares owned by the largest shareholders. SMALL s defned as the percentage of loans for medum and small frms / total loans and blls dscounted. TAR[Number] and TAR[Yen] are defned as the percentages of the announced target shares and yen value for repurchase, respectvely. TAR[Number] represents the percentage of the number of shares targeted for repurchase to the total number of outstandng shares. TAR[Yen] represents the percentage of the yen value of shares targeted for repurchase to the market value of outstandng shares as of March 00. EXR[Number] and EXR[Yen] are defned as the percentage of the actually repurchased shares and yen value up to the date of the next shareholder meetng. EXR[Number] represents the percentage of the number of actual repurchased shares to the number of the target shares. EXR[Yen] represents the percentage of the yen value of shares repurchased actually to the yen value targeted for repurchase. 19

21 Table. Descrptve Statstcs Concernng Cash Dvdends and Share Repurchases: The Perod form 000 and 00 Repurchase Group [RG] (N=45) Non-Repurchase Group [NRG] (N=40) Panel (1) Number of Banks () Dvdend payments () Interm dvdend payments Ether () or () Panel () Dvdends(per a unt of share) Mean Medan Standard Devatons Panel (3) Correlaton Between Executed amount and Dvdend payments 0.54 *** Note: ***, **, and * present sgnfcant at the 1%, 5%, and 10% levels, respectvely. 0

22 Table 3. Tobt Estmates Predctng the Target for Share Repurchases The table provdes estmates of the relaton between the target for share repurchases and varables lkely to affect the repurchase decson. Our dependent varables are TAR[Number] and TAR[Yen]. These are censored varables of whch the value are larger than zero f the resoluton regardng the target for share repurchases s made by the board of drectors, zero otherwse. We estmate the one-taled Tobt model truncated at 0%. As explanatory varables, we utlze the constant term, the logarthm of SIZE, whch s defned as the book value of Total Assets mnus Equty plus the market value of Equty, the market to book value rato M/B, the excess captal to asset rato ECR, whch s defned as the rsk-based captal rato reported n bank s fnancal statement less the requred captal rato, and the rato of non-performng loans to total loans NPL/L. Sgma denotes the standard devaton of the dsturbances n the Tobt model. Dependent Varable = TAR[Number] Dependent Varable = TAR[Yen] Independent Varables Panel (1) Panel () Panel (3) Panel (4) Coeffcents t-ratos Coeffcents t-ratos Coeffcents t-ratos Coeffcents t-ratos ln(size) *** ** *** ** ROA M/B *** ** *** ** ECR NPL *** *** Sgma *** *** *** *** Pseudo-R Notes: ***, **, and * present sgnfcant at the 1%, 5%, and 10% levels, respectvely. The estmate of the ntercept term s omtted for want of space. 1

23 Table 4. Tobt Estmates Predctng the Target for Share Repurchases The table provdes estmates of the relaton between the target for share repurchases and varables lkely to affect the repurchase decson. Our dependent varables are TAR[Number] and TAR[Yen]. These are censored varables of whch the value are larger than zero f the resoluton regardng the target for share repurchases s made by the board of drectors, zero otherwse. We estmate the one-taled Tobt model truncated at 0%. We append the dfference of SHARE, SHARE, to the vector of explanatory varables. Dependent Varable = TAR[Number] Dependent Varable = TAR[Yen] Independent Varables Panel (1) Panel () Coeffcents t-ratos Coeffcents t-ratos ln(size) ** ** ROA M/B * * ECR NPL *** *** SHARE Sgma *** *** Pseudo-R Notes: ***, **, and * present sgnfcant at the 1%, 5%, and 10% levels, respectvely. All equatons nclude the ntercept term.

24 Table 5. Tobt Estmates Predctng the Executon of Share Repurchases Ths table provdes estmates of the relaton between the executon of share repurchases and varables lkely to affect the repurchase decson. We use 45 share-repurchasng banks, whch have announced ther repurchase program at the shareholder s meetng n the fscal year of 00, to construct the two percentages of shares that frms actually repurchase; EXR[Number] and EXR[Yen]. These are censored varables of whch the value are larger than zero f banks actually repurchase ther own shares, zero otherwse. There are no banks of whch the executon rato reaches 100% wthn one year. We, therefore, estmate the one-taled Tobt model truncated at 0%. As the explanatory varables affectng bank s executon of share repurchases, we use the constant term, LCAR, ln(size), M/B, ECR, and NL/L. In the no executon group, LCAR s defned as the cumulatve abnormal return from the date of the shareholder s meetng to the date when banks actually repurchased ther own shares for the frst tme. On the other hand, n the no executon group, LCAR s defned as the cumulatve abnormal return from the date of the shareholder s meetng to the date of the next shareholder s meetng. The explanatory varables M/B, ECR, and NL/L are based on a bank s fnancal statements n 003. Therefore, allowng for the endogenety of the explanatory varables M/B, ECR, and NL/L, we estmate the Tobt model wth endogenous varables by maxmum lkelhood estmaton. We utlze the constant term, LCAR, and the one-lagged and two-lagged values of ln(size), M/B, ECR, NL/L, and SHARE as nstrumental varables n estmatng our model. Sgma denotes the standard devaton of the dsturbance n the Tobt model. Independent Varables Dependent Varable = TAR[Number] Dependent Varable = TAR[Yen] Coeffcents t-ratos Coeffcents t-ratos LCAR *** *** ln(size) M/B ECR * ** NPL ** ** Sgma *** *** Pseudo-R Notes: ***, **, and * present sgnfcant at the 1%, 5%, and 10% levels, respectvely. All equatons nclude the ntercept term. 3

25 Table 6. Tobt Estmates Predctng the Executon of Share Repurchases Ths table provdes estmates of the relaton between the executon of share repurchases and varables lkely to affect the repurchase decson. We use 45 share-repurchasng banks, whch have announced ther repurchase program at the shareholder s meetng n the fscal year of 00, to construct the two percentages of shares that frms actually repurchase; EXR[Number] and EXR[Yen]. These are censored varables of whch the value are larger than zero f banks actually repurchase ther own shares, zero otherwse. There are no banks of whch the executon rato reaches 100% wthn one year. We, therefore, estmate the one-taled Tobt model truncated at 0%. As the explanatory varables affectng bank s executon of share repurchases, we use the constant term, LCAR, ln(size), M/B, ECR, NL/L, and SHARE. In the no executon group, LCAR s defned as the cumulatve abnormal return from the date of the shareholder s meetng to the date when banks actually repurchased ther own shares for the frst tme. On the other hand, n the no executon group, LCAR s defned as the cumulatve abnormal return from the date of the shareholder s meetng to the date of the next shareholder s meetng. The explanatory varables M/B, ECR, NL/L, and SHARE are based on a bank s fnancal statements n 003. Therefore, allowng for the endogenety of the explanatory varables M/B, ECR, NL/L, and SHARE, we estmate the Tobt model wth endogenous varables by maxmum lkelhood estmaton. We utlze the constant term, LCAR, and the one-lagged and two-lagged values of ln(size), M/B, ECR, NL/L, and SHARE as nstrumental varables n estmatng our model. Sgma denotes the standard devaton of the dsturbance n the Tobt model. Independent Varables Dependent Varable = TAR[Number] Dependent Varable = TAR[Yen] Coeffcents t-ratos Coeffcents t-ratos LCAR *** *** ln(size) M/B ECR * ** NPL ** ** SHARE Sgma *** *** Pseudo-R Notes: ***, **, and * present sgnfcant at the 1%, 5%, and 10% levels, respectvely. All equatons nclude the ntercept term. 4