WIF-10-001 February 2010 Ownership srucure, liquidiy, and firm value Jun Uno and Naoki Kamiyama 1
Ownership srucure, liquidiy, and firm value February 2010 Jun Uno * and Naoki Kamiyama * Waseda Universiy, Graduae School of Finance, Accouning and Law.This research is suppored by he Financial Service and Innovaion Managemen Research Projec, sponsored by he Minisry of Educaion, Culure, Spors, Science and Technology, FY2008. Correspondence o Jun Uno, 1-4-1 Nihombashi, Chuo-ku, Tokyo 103-0027, Japan; Tel. 81-3-3272-6798; fax 81-3-3272-6783; e-mail juno@waseda.jp. Deusche Securiies Inc.
Absrac A firm s ownership srucure influences is liquidiy and value. In addiion o adverse selecion coss, shareholders invesmen horizons and syles are key deerminans in his relaion. This paper inroduces a new measure of laen invesmen horizon, a weighed average invesmen horizon compued from he firm s ownership srucure and he average invesmen horizon of various invesor caegories. We find ha boh invesmen horizon and adverse selecion cos have negaive effecs on liquidiy, bu opposie effecs on firm value among firms lised on he Tokyo Sock Exchange. Empirical resuls sugges ha he longer he invesmen horizon, he lower he firm s liquidiy and value. Concenraed ownership by he op 30 shareholders can increase firm value, bu cross-holding shareholders decrease i. JEL classificaion: G10, G32 Keywords: Ownership srucure, liquidiy, invesmen horizon, adverse selecion, invesmen syle, marke microsrucure, monioring, governance 2
I. Inroducion This paper invesigaes how ownership srucure is relaed o a firm s marke liquidiy and value. If he marke liquidiy of a firm s shares declines due o concenraed ownership, he firm s value is expeced o decrease. Illiquidiy can resul from increased asymmeric informaion. On he oher hand, monioring by a large shareholder srenghens corporae governance, increasing a firm s value. Therefore, here may be a rade-off beween he cos of illiquidiy and he meris of corporae governance (Bhide 1993, Holmsrom and Tirole 1993). Kahn and Winon (1998) sugges ha greaer liquidiy can be an opporuniy for large shareholders o increase heir profis by monioring a firm s managemen. The auhors menion he case where a large shareholder chooses o buy more shares when he firm s performance is expeced o improve as a resul of monioring aciviies. The greaer he liquidiy, he more shares can be bough in he marke due o lower ransacion coss. Thus, a higher concenraion of ownership does no necessarily mean a rade-off beween corporae governance and liquidiy. This finding differs from pas views, such as in Bhide (1993), where a sock s high liquidiy renders large shareholders less aggressive in heir monioring and more likely o sell shares. Prior heoreical research, however, does no consider he effec of he shareholder invesmen horizon on a firm s marke liquidiy and value. If a large proporion of a firm s shareholders have a longer invesmen horizon, he firm s marke liquidiy diminishes. On he oher hand, if a firm has many invesors wih a shorer invesmen horizon, is liquidiy improves. We show ha he average 3
invesmen horizon of he owners is a key deerminan of a firm s liquidiy and value, as Amihud and Mendelson (1986) predic. Rubin (2007) empirically examines he effec on liquidiy under he adverse selecion and rading hypoheses. The adverse selecion hypohesis refers o he case where illiquidiy can resul from increased asymmeric informaion. The rading hypohesis addresses he case where a large proporion of a firm is owned by insiuional invesors and heir rading aciviy improves liquidiy. This sudy breaks down he rading hypohesis ino wo elemens, invesmen horizon and invesmen syle. The firs elemen relaes o rading frequency, and he second one o order imbalance. We use hree elemens o explain he relaions beween a firm s ownership srucure and is liquidiy and value. We esimae he laen invesmen horizon of he shareholders of Japanese companies lised in he Tokyo Sock Exchange (TSE). Following Mahani e al. (2008), who esimae he invesmen horizon of corporae bondholders using daa from cusodian banks, we compue a weighed average invesmen horizon from he equiy ownership srucure of he TSE-lised companies. In our case, however, here are no corresponding daa available from cusodian banks. Therefore, we esimae hese from wo sources of daa. Firs, each invesor group s average invesmen horizon is esimaed from he aggregae amoun of he holding and rading volume of all he lised companies. This provides marke-wide average invesmen horizons by invesor caegory. Second, we obain individual firm ownership raios in each invesor caegory from he firms annual repors. Using 4
marke-average invesmen horizons and firm ownership raios, we hen calculae each firm s weighed average invesmen horizon. Thus our measure is differen from ha compued from he firm share urnover raio used in prior research, such as Akins and Dyl (1997). These auhors examine he relaion beween invesors average holding periods and liquidiy in a manner similar o ours. They use share urnover as a proxy for shareholder invesmen horizon. In heir sudy, he urnover-based invesmen horizon is no direcly linked o ownership srucure because rading volume can vary for many reasons. For insance, quarerly earning announcemens and insiuional porfolio rebalancing increase rading volume. On he oher hand, rading volume is affeced by a kind of day-rading ha is unrelaed o ownership srucure. Since his paper focuses on how ownership srucure affecs sock liquidiy, we capure he characerisics of firm shareholder composiion. By considering he weighed average invesmen horizon of shareholders, we can address cases in he real economy where, under he same concenraed ownership, a firm s marke liquidiy is higher because he invesmen horizon of is shareholders is shorer han ohers. We show ha a firm s invesmen horizon is highly correlaed wih marke liquidiy and rading cos. We expec ha he shorer he invesmen horizon, he higher he liquidiy. Addiionally, here are wo disinc caegories of large shareholders in Japan: foreigners and cross-holders. These invesors are opposies in erms of invesmen horizon, syle, and governance. We shed some ligh on specific invesor caegories and wheher he size of heir presence affecs firm liquidiy as well as value. 5
Our resuls are summarized as follows: (1) The shorer he laen invesmen horizon, he higher he liquidiy and firm value. (2) The level of ownership concenraion is negaively relaed o liquidiy, bu posiively relaed o value. I suggess ha he posiive effec on firm value brough abou by beer governance exceeds he negaive effec imposed by illiquidiy. (3) Foreigners, an invesor group wih well-diversified invesmen syles, have a posiive effec on liquidiy and value and herefore a posiive impac. Cross-holders, he leas acive and mos homogeneous invesor group, have a negaive effec on liquidiy and value. These resuls imply ha hree elemens of ownership srucure relae o a firm s liquidiy and value. Our resuls indicae ha he higher he proporion of shor horizon shareholders, he higher a firm s value. On he oher hand, a firm wih a high proporion of infrequen invesors resuls in lower firm liquidiy and value. Based upon hese resuls, encouraging cross-holding is an inferior corporae policy, since i ulimaely impairs he liquidiy of he enire marke. The remainder of his paper is organized as follows. Secion II compares our new approach o he exising lieraure. The daa and sample socks are described in Secion III. Secion IV presens our empirical resuls and Secion V draws our conclusions. II. Ownership, Liquidiy, and Value Firm ownership and is effecs on marke liquidiy were originally discussed in corporae finance and corporae governance lieraure (Bhide 1993, Holmsrom and Tirole 1993, Kahn and Winon 1998, Maug 1998). Liquidiy, measured by 6
bid ask spread, has been sudied from he perspecive of marke maker (or invesor) behavior and/or marke design (Glosen and Milgrom 1985, Glosen and Harris 1988, Huang and Soll 1997). Recenly he relaion beween firm ownership and liquidiy has been recognized as an imporan bridge beween corporae finance and marke microsrucure (Rubin 2007, Amihud and Mendelson 2008). Wih regard o adverse selecion, Bhide (1993) and Holmsrom and Tirole (1993) demonsrae a negaive correlaion beween ownership concenraion and liquidiy. The informaional advanage of large shareholders who commi o monioring managemen increases informaion asymmery among invesors, lowering liquidiy. Kahn and Winon (1998) and Maug (1998) sugges ha large shareholders increase heir ownership when marke liquidiy is high, he decision of how many shares o add depending upon marke liquidiy. The auhors also poin ou ha greaer accuracy of large shareholders informaion increases rading profis. The greaer he liquidiy, he more shares can be bough in he marke due o lower ransacion coss. Thus, a higher concenraion of ownership does no necessarily cause illiquidiy due o adverse selecion coss. This rading aspec of shareholders relaes o Amihud and Mendelson (1986), whose model predics ha illiquid socks are owned by invesors wih longer invesmen horizons. The relaion beween liquidiy and shareholder behavior depends upon shareholder invesmen horizons as well as invesmen syle. A small number of empirical sudies have been carried ou on he relaion beween ownership srucure and liquidiy. For example, Akins and Dyl (1997) sudy he 7
relaion beween invesor average holding periods and he bid ask spread for NASDAQ and New York Sock Exchange socks. The auhors find srong evidence ha he share urnover raio, compued by dividing a firm s number of shares ousanding by is annual rading volume, is negaively correlaed wih he bid ask spread. They do no, however, consider ownership srucure or is direc relaion o share liquidiy. Rubin (2007) finds ha liquidiy is posiively correlaed o oal insiuional holdings bu negaively correlaed o insiuional block holdings. The level of insiuional holdings proxies for rading aciviy, and he concenraion of ownership, such as block holdings, proxies for adverse selecion coss. Sarin e al. (2000) analyze ownership concenraion by insiders and by insiuional invesors. The auhors repor decreased liquidiy in boh cases: by insiders from increased asymmeric informaion and by insiuional invesors from invenory coss. Foreigners ypically have a shorer invesmen horizon and are ap o monior managemen, whereas cross-holding shareholders ypically have a longer invesmen horizon and are less acive in monioring managemen. According o Nia (2000) analysis of daa from 1988 o 1997, a posiive correlaion exiss beween he raio of foreign ownership and managemen performance measures such as reurn on equiy (ROE), bu a negaive correlaion exiss in he case of he cross-holding ownership raio and managemen performance. We expec ha he larger he percenage of foreign ownership, he higher he liquidiy; on he oher hand, he larger he percenage of cross-holding owners, he lower he liquidiy. 8
The role of foreign invesors has been focused upon from a home bias poin of view. Tesar and Werner (1995) find ha he urnover rae on equiy held by non-residens is higher han he overall urnover rae on he domesic marke. Foreigners respond o changes in economic condiions by making frequen and sizable shifs in heir holdings of foreign securiies, even hough much of his aciviy has lile impac on ne invesmen posiion. III. The Daa We use four years of daa on Firs- and Second-Secion companies on he TSE, from 2004 o 2007. The sample includes 1,657 o 1,686 socks for which liquidiy and cross-holding daa are fully available. A. Ownership concenraion Our measure of ownership concenraion is he concenraion raio of he op 30 shareholders, which is he sum of he op 30 shareholders holdings divided by he oal number of shares ousanding. This measure serves as a proxy for he probabiliy ha informed non-insider invesors paricipae in he marke. The concenraion of he op 30 ownership srenghens he severiy of adverse selecion coss. Rubin (2007) uses insider and insiuional block holdings o measure concenraion. Harzell and Sarks (2003) use a measure ha is he sum of he op five non-insider insiuional invesors holdings divided by he oal number of insiuional holdings. We adjus our measure in case insider block holders are included in he op 30 shareholders. 9
B. Invesmen horizon A firm s invesmen horizon is esimaed as follows. The invesmen horizons of an invesor group are firs compued from he aggregaed marke daa provided by he TSE. Each invesor group s invesmen horizon is he oal rading volume during he year divided by he average porfolio marke value a he sar and end of each year. The daa source for sock ownership by invesor group is he TSE s Share Ownership Survey. The rading amoun of each invesor group is compiled by he TSE and published annually in is Invesmen Trends by Invesor Group. 1 There are four invesor caegories used a his sage of he esimaion; foreigners, individuals, non-financial corporaions, and a group of financial insiuions (rus banks, insurance companies, and banks). 2 The following is he equaion for an invesor group j s urnover rae in year : j j 1 ( porfolio marke value porfolio marke value 1) j (1) InvesmenHorizon 2 j oal rading volume (Table I around here) Table I shows he average invesmen horizon for each invesor group a he aggregae level. A non-financial corporaion s invesmen horizon (Horizon) is 5 o 7 years and an insurance company s Horizon is 15 o 21 years, which is he mos 1 The original monhly daa frequency was convered ino fiscal years. The ownership survey covers seven local markes in Japan, whereas rading volume covers hree major markes. We believe his discrepancy has very lile impac on our resuls. 2 In he classificaion of financial insiuions, he invesor caegories in annual repors do no mach he hree subgroups available in he TSE saisics, such as commercial banks, insurance companies, and rus banks. Therefore, we ake an average of he invesmen horizons for commercial banks, insurance companies, and rus banks as he invesmen horizon for financial insiuions when compuing he invesmen horizons of he TSE-lised companies. 10
inacive among six invesor caegories in Table I. The Horizon of foreigners is 0.2 o 0.4 year, which means hey rade wo o five imes a year. The Horizon of individuals is 0.3 o 0.5 year. The finding ha foreigners Horizon is shorer han ha of domesic invesors is consisen wih Tesar and Werner (1995). The auhors repor urnover raes in domesic equiy held by foreign residens for five major counries including he Unied Saes, he Unied Kingdom, and Japan and find ha foreign invesors ransac a a significanly higher rae han domesic invesors. Horizon for firm k in year is compued as k (2) Horizon w 4 j 1 j k, Horizon j where j represens one of four invesor caegories, foreigners, individuals, non-financial corporaions, and financial insiuions; j w k, is a firm j s ownership raio obained from he company s annual repor; and k Horizon is he marke-wide invesmen horizon by invesor group. The daa source for ownership srucure was Quick s AMSUS. Thus, he invesmen horizons for all lised firms are esimaed, since hey reflec differences in ownership srucure. The following example illusraes he previous compuaion. In 2007 Sony Corporaion s ownership raios j w k, of foreigners, individuals, non-financial corporaions, and financial insiuions were 50.73, 23.11, 3.1, and 21.57%, 11
j j respecively. Using he Horizon in Table I and k, we compue he weighed average of Sony s invesmen horizon, which is abou 2.8 years: w, (3) A weighed average of Horizon 1 0.217 50.73% 0.373 23.11% 6.77 3.1% (0.959 16.659 15.650) 21.57% 3 0.11008 0.08620 0.20987 2.39197 2.8 year We used he NEEDS daabase as an addiional source of daa on large individual shareholders such as he founder (owner) of a firm and naional and local governmens on he op 30 shareholders lis. We use hese daa o adjus he above compuaion. Since neiher a founder (owner) nor a governmen rade like ordinary individual invesors, we assign he longes invesmen horizon among six subinvesor caegories in he same year o heir ownership caegory. C. Invesor group ownership raio The ownership raio is compued as he number of shares owned by each invesor group divided by he number of shares ousanding. Foreign and individual invesors are used in a regression analysis. The variable for Individuals holdings is highly correlaed wih Horizon and Top30, such ha i creaes a mulicollineariy problem. We regress Horizon and Top30 on Individuals holdings. (4) Individual s holdings a Horizon 30 1 2Top We obain he variable Indiv used in he Secion IV, which is he residual erm of he equaion (4). 12
D. Cross-holding Our measure of so-called cross-holding shareholders is he Mochiai holding raio of each sock, provided by he NLI Research Insiue. The Mochiai holding occurs when wo lised companies muually hold shares, confirmed hrough disclosed maerials. 3 An average mochiai raio is around 9% and he maximum is 50 o 55%. 4 E. Liquidiy The proxy of marke illiquidiy is Amihud s (2002) ILLIQ, he monhly average of he absolue daily reurn divided by he daily yen volume. We eliminae socks raded less han 10 days per monh. Here, ILLIQ shows he relaion beween price change and volume; i is a rough esimae of spread plus marke impac cos. We use he average relaive illiquidiy (RILLIQ) for each fiscal year: (5) (6) (7) ILLIQ iy 1 D D iy iy 1 N y 1 R iyd VOLD ivyd 1 AILLIQ y ILLIQ iy N RILLIQ iy ILLIQ iy AILLIQ y Here, RILLIQ is adjused for marke-wide liquidiy changes and convered ino a naural logarihm. 5 According o he lieraure, liquidiy has a commonaliy 3 Alernaive measure is he Anei holding, where he ownership of a firm s shares by banks and life insurance companies is disclosed bu he firm s holdings of counerpary shares canno be confirmed by disclosed maerials. Refer o Nia (2002) for deails. 4 Nia (2002) shows ha he higher he cross-holding raio, he worse he various managerial indices, such as he ROE. 5 The impac of a change in marke-wide liquidiy see equaion (2) on ILLIQ for individual socks is no uniform: I disproporionaely affecs low-liquidiy socks (see Amihud 2002 and Chordia e al. 2000). 13
(Chordia e al. 2000), and is marke-wide flucuaion has differenial impacs on he price of individual socks (Amihud 2002). Thus, RILLIQ is adjused for a ime series variaion of marke-wide liquidiy changes. F. Q raios Our proxy for firm value is he surrogae Tobin s q = (aggregae marke value + ineres-bearing deb)/(oal capial + ineres-bearing deb). The financial daa for calculaing his are from Worldscope. When we compue Tobin s q, he cross-holding-adjused aggregae marke value is used. 6 Table II shows he summary saisics of hese variables. G. Correlaions Table IIIA shows he correlaions among he variables. As shown in Table IIIA, Horizon is posiively correlaed wih he logarihm of marke capializaion (0.3) and negaively correlaed wih he share urnover raio (0.11). The correlaion wih he cross-holding ownership raios of Crosshld is posiive (0.3), meaning ha he greaer he value of Horizon, he higher he cross-holding raio. On he oher hand, here is almos no correlaion beween Horizon and Foreigner (see Table IIIB). (Table II and Table III around here) 6 Kobayashi (1990) poins ou ha he Mochiai porion should be subraced from he aggregae marke value o calculae Tobin's q. 14
IV. Empirical Analyses A. Hypoheses Firs, we examine he relaion of ownership srucure wih liquidiy by esing he following four hypoheses. (1) Adverse selecion hypohesis: The higher he ownership concenraion, he lower he liquidiy. The variable for ownership concenraion is Top30 which is our proxy for he severiy of adverse selecion coss. (2) Invesmen horizon hypohesis: The longer he invesmen horizon, he lower he liquidiy. If a firm has many long-erm invesors, hey will rade infrequenly such ha he liquidiy of he sock will remain low. The variable for invesmen horizon is Horizon. (3) Invesmen syle hypohesis: The larger he share of foreign ownership, he higher he liquidiy. The larger he ownership share of individuals, he lower he liquidiy. Based upon he resuls of prior research, we assume ha invesor groups such as foreigners and individuals each have heir own characerisic invesmen syle. I has been recognized ha he invesmen syle of foreigners is diversified, and herefore a greaer paricipaion of foreign invesors can miigae order imbalance and lower rading coss. Individuals are viewed as uninformed invesors who can miigae he cos of adverse selecion, so ha heir large presence improves liquidiy. The variables for foreign invesors and individual invesors are Foreign and Indiv, respecively. 15
(4) Governance hypohesis: The larger he number of cross-holding shares, he lower he liquidiy. A large proporion of cross-holding shareholders means ha proecion for minoriy shareholders is weak, providing evidence ha governance saus affecs he liquidiy of a firm s shares. The variable for cross-holding shares is Crosshld. The dependen variable is RILLIQ, our proxy for marke illiquidiy. The naural logarihm of a firm s marke value (Log_Size) and ADRdummy 7 are added as conrol variables. 8 The basic regression equaion is (8) RILLIQ 5 1 2 30 Crosshld a b Horizon 6 Log_Size b Top 7 3 ADRdummy Foreign 4 Indiv B Endogeneiy Liquidiy and ownership srucure may be simulaneously deermined. In equaion (8), all variables are mos likely endogenous and he esimaes based on panel leas squares are biased and inconsisen. Considering hese problems, we use a wo-sage esimaion mehod wih insrumenal variables o obain a consisen esimae. 9 Our insrumenal variables are he lagged explanaory variables. For Horizon and Top30, we also es alernaive specificaions for 7 When a company issues American Deposiary Receips (ADRs), ADRdummy = 1; oherwise ADRdummy = 0. Kang and Sulz (1997) sugges ha ADR is an imporan variable when considering foreign equiy ownership. 8 Amihud (2002) shows a high negaive correlaion (-0.614) beween size and ILLIQ. In our case he correlaion is -0.57. 9 Woodridge (2002) Chaper 10-11. 16
nonlineariy. Heeroskedasiciy is correced by Whie diagonal sandard errors and covariance correcions. Firs, we regress four variables (Top30, Horizon, Log_Size, and ADR_dummy) on RILLIQ. In Table IV, model1 shows he resul of a linear relaion beween Horizon and Top30. The coefficien of Top30 is posiive and significan a he 1% confidence level, which means ha he higher he Top30 concenraion, he lower he liquidiy. This finding is consisen wih he adverse selecion cos hypohesis. The coefficien of Horizon is also posiive and significan a he 1% level. The longer he invesmen horizon, he lower he liquidiy. This finding is consisen wih he second hypohesis, ha he invesmen horizons are associaed wih he level of illiquidiy. Model2 examines he nonlinear relaion of he wo variables Top30 and Horizon. The resuls are almos idenical, and he -values of Top30 and Horizon are slighly higher han hose in model1. The adjused R-squared values do no change much. We ried o incorporae boh a linear and a quadraic erm of he wo variables, no repored in Table IV, bu boh variables become insignifican. These resuls sugges ha nonlineariy is no beer specificaion han lineariy for he wo variables. We add hree invesor group variables (Foreign, Indiv, and Crosshld) in model3 and model4 o represen invesor caegories. As expeced, Foreign and Indiv have saisically significan negaive relaions wih RILLIQ, bu Crosshld has a significan posiive relaion. These findings mean ha he higher he proporions of foreign and individuals ownership, he greaer he liquidiy, while he higher 17
he proporion of cross-holding ownership, he lower he liquidiy. This finding is consisen wih he hird hypohesis. Foreign invesors engage in a variey of invesmen sraegies, such as value, growh, and momenum. Hochkiss and Srickland (2003) repor on he heerogeneiy in U.S. insiuional rading behavior and noe ha he magniude of price reacions a earnings announcemens is relaed o no only he level of insiuional ownership bu also he characerisics of he insiuional holders. Amihud e al. (1999) show ha an increase in he individual invesors helps improve liquidiy. C. Selecion bias Ownership daa such as Foreign are a ype of censored daa ha are zero when an invesor does no inves in a sock. The decision o inves or no can be based upon consideraions oher han he number of shares invesed. Considering hese problems, we use he Heckman wo-sep mehod o obain a consisen esimae. 10 The firs sep in he esimaion of he probi selecion model for Foreign and Crosshld uses all observaions o obain he inverse Mills raio, which is he bias correcion erm for he second sep. Foreign invesors prefer large, liquid socks ha are expor-oriened global companies. We selec he firm s marke value, overseas sales, and he ADR dummy a dummy variable ha equals 1 if a firm has an ADR program and 0 oherwise as deerminans for Foreign. 11 The ownership raios we use in his sudy are he aggregaed numbers of shares owned by each invesor group. The number of socks no held by foreigners is only 10 Woodridge (2002) Chaper 10-11.. 11 Kang and Sulz (1997) sugges ha overseas sales and an ADR program are firm aribues ha migh affec foreign invesors preferences. Dahlquis and Robersson (2001) suggess ha a lis of facors relaed o foreign invesors preferences. 18
12 ou of 6,695 samples due o insufficien daa. This sample size is oo small o esimae a probi selecion model. Foreigners, as a group, owning less han 1% of a firm can be viewed as having nearly zero ownership. Their presence does no have any effec on liquidiy or a firm s value. Thus, we draw he line beween holding and quasi-no-holding a 1% of a firm s shares ousanding, caegorizing 647 socks as quasi-no-holding and 6,048 socks as holding. This change improves he esimaion condiions for he probi selecion model. We also draw he line a 10% of he firm s shares ousanding. When foreign invesors become concerned wih monioring managemen, heir incenives can be srenghened by a higher ownership raio. We choose 10% as an alernaive hreshold for he foreigners probi model among he meaningful ownership raios, such as 1, 3, or 10%, according o shareholder righs graned by Japanese corporae law. 12 The 10% hreshold breaks up he enire sample ino approximaely equal pars. The selecion models for foreign ownership 1% (10%) are (9) Forein _1%(10%) 4 ROE Growh 5 a Log_Size 1 Horizon 2 Expor 3 ADRdummy The selecion model for crossholding is (10) Crosshld _ dummy Growh 4 a Log_Size Floa 5 1 Horizon 2 Expor 3 ADRdummy 12 A shareholder wih 1% ownership can raise a proposal a he shareholders meeing, one wih 3% ownership can call for an exraordinary shareholders meeing, and one wih 10% has a claim of dissoluion by sockholders. These righs can be exercised collecively. 19
The resuls of he foreigner selecion models are shown in Table V. In he case of a 1% hreshold (model1), he explanaory variables of Log_Size and Expor are posiive and saisically significan. The larger a company s size and he greaer is overseas sales, he more likely foreign invesors are o own more han 1% of is shares ousanding. The oher hree variables (ADRdummy, ROE, and Growh) are insignifican. In he case of a 10% hreshold (model2), all he variables have posiive, significan coefficiens excep Growh, which has a negaive, significan coefficien. In addiion o size and expor, foreigners prefer issuers of ADRs and high-roe firms. We also esimae he selecion model for cross-holding shareholders. The number of socks ha cross-holding shareholders do no hold is 715 ou of 6,695 samples. The cross-holding model can be esimaed wihou adjusing he hreshold. Here, Floa is added as an explanaory variable, which is he number of shares no held by managemen or cross-holders. The higher he value of Floa, he greaer he chance of being a arge for corporae raiders. Therefore, managers care abou srenghening heir cross-holding relaionships. Oher variables such as size and overseas sales are relevan for hose firms ha encourage cross-holdings as proecion agains poenial akeover acion. 13 The resuls of he cross-holding selecion models are shown as model3 in Table V. The explanaory variables Log_Size, Expor, and Floa have posiive and saisically significan relaions wih he level of cross-holding. The larger he size and he greaer overseas sales, he higher he cross-holding raio. Here, ADR 13 The resuls of he probi selecion model are available upon reques. In his version, he resuls are aached as a supplemenary able a he end of he paper. 20
has a negaive relaion wih cross-holding because firms who have ADR programs may esablish good governance; hey usually do no need srenghening cross-holding o be proecion agains akeovers. The signs of he explanaory variables are as expeced. The second sep runs he following equaion, which includes he inverse Mills raios (imills) of foreigners and cross-holders: (11) RLLIQ 5 Indiv a Horizon ADRdummy 9 Top30 imillsforeign imillscrosshld Log _ Size 6 1 2 7 3 Foreign 8 4 Crosshld where variables such as RILLIQ, Log_Size, Horizon, Foreign, Indiv, and Crosshld are same as in equaion (8). We esimae equaion (11) by he wo-sage panel leas squares mehod wih lagged explanaory variables as insrumenal variables. Heeroskedasiciy is correced by Whie diagonal sandard errors and covariance correcions. Table VI shows he esimaion resuls of equaion (11), where he resul for he invesmen horizon is unchanged from Table IV. In he case of a 1% hreshold for he foreign ownership raio (model1), he inverse Mills raio is negaive and significan a he 1% level. This means ha foreigners choices o inves in a sock are linked o ha sock s liquidiy. A lower RILLIQ or greaer liquidiy is associaed wih he decision of foreigners o inves in he sock. In he case of a 10% hreshold (model2), he inverse Mills raio is also negaive and significan a he 1% level and is coefficien is much smaller (-0.2139) han for he 1% hreshold 21
(-1.24). This suggess a concave relaion beween liquidiy and foreigners invesmen decisions. I is imporan o noe ha he coefficien of Foreigners is sill negaive and significan afer accouning for he endogeneiy problem. This means ha greaer invesmens by foreigners lead o lower RILLIQ, or higher liquidiy. This resul relaes o Hochkiss and Srickland (2003), who examine he impac of ideniy and rading behavior and provide an imporan view on insiuional order flow ha is affeced by a broad diversiy of invesing syles. Orders from insiuions can offse each oher o reduce price impacs. Thus, diversified invesmen syles and he shor-erm rading behavior of foreign invesors conribue o lowering ransacion coss. The resuls for Indiv are opposie o hose in Table IV. Afer including he inverse Mills raio, he coefficien of Indiv urns posiive, which means he greaer Indiv, he lower he liquidiy for boh he 1 and 10% hresholds. The effecs of Indiv appear o be unsable, alhough we remedy is high correlaion wih Top30 and Horizon. The inverse Mills raio from he cross-holding selecion model is posiive and significan a he 1% level. This means ha he choice made by cross-holding invesors o hold shares is linked o he sock s greaer illiquidiy. Higher RILLIQ, or greaer illiquidiy, is associaed wih he choice of cross-holding shareholders o inves in he sock. Typical cross-holding invesors are banks, whose average invesmen horizon is 11 o 17 years. The coefficien of he level of cross-holding is 22
negaive and significan. This finding means ha he percenage ownership of cross-holding also has an impac on illiquidiy. The resuls in Table VI for foreign and cross-holders are consisen wih Kahn and Winon s (1998) predicion: When a firm s large shareholders commi o monioring managemen, hey frequenly rade on heir privae informaion o maximize heir profis and hus enhance he sock s liquidiy. On he oher hand, when large shareholders do no acively monior firm managemen, hey do no rade so much ha he firm s liquidiy decreases. (Table VI around here) D. Robusness check As an addiional robusness check, he relaion beween he frequency of rades and he average invesmen horizon is examined. A direc implicaion of shor ime horizons is ha shareholders of he sock rade more ofen. To verify he robusness of he horizon as a liquidiy facor, we use Log_Trade insead of RILLIQ o run regressions similar o hose in equaions (8), where Log_Trade is he naural logarihm of he number of rades in year for firm j. We expec ha he longer he horizon, he less frequen he rades. We also expec a negaive correlaion beween he number of rades and he ownership concenraion raio. Here, he Top30 ownership raio indicaes he high level of adverse selecion risk. (12) Log _ Trade a Horizon Top30 (% Sprd ) Log Size 1 2 3 _ 23
where %Sprd, is he average percenage bid-ask spread in year for firm j. Heeroskedasiciy is correced by Whie diagonal sandard errors and covariance correcions. The resuls of he regression presened in Table VII are as expeced. The smaller he value for Horizon, he larger ha for Log_Trade, where he coefficien of Horizon is negaive and significan a he 1% level. The coefficien of Top30 is also negaive and significan a he 1% level. This finding means ha he lower he ownership concenraion, he larger Log_Trade. The sign and significance of Horizon do no change afer adding %Sprd as an explanaory variable. However, he coefficien of Top30 urns ou o be insignifican due o he high correlaion beween Spread and Top30. Higher Spread is associaed wih higher adverse selecion risk. (Table VII around here) E. Ownership and liquidiy and heir effecs on firm value We esimae he influence of invesmen horizon and ownership concenraion on firm value. According o he posiive correlaion beween invesmen horizon and illiquidiy found in able VI, we expec ha he shorer he invesmen horizon, he higher he firm value. In equaion (13), Top30 and Crosshld are variables ha relae o governance. 14 Regarding he ownership concenraion, here is a rade-off beween he coss of 14 Morck e al. (2000) repor ha he equiy ownership by corporae block holders is posiively relaed o firm value in Japan. They use a sample of 373 manufacuring firms lised on he Firs secion of he TSE in 1986. 24
illiquidiy and he meris of corporae governance. We will see which effec is dominan. In case of cross-holding, we assume ha he lower Crosshld, he beer he governance. Tobin s q (QRaio) is used as a proxy for he firm s marke valuaion. We esimae equaion (10) by he Heckman wo-sep mehod. As in equaion (10), imills from Foreign and Crosshld obained from he probi selecion model are included o correc bias from he selecion. The second-sage regression equaion (13) is QRaio a Horizon imillsforeign _1% 7 1 Top30 immilscrosshld 8 2 Foreign 4 Indiv Debasse 9 5 Crosshld 10 6 Growh Log _ Size where he deb asse raio (Debasse), Growh and Log_Size are included as conrol variables. Here, Debasse is calculaed as oal deb divided by oal asses, and Growh is calculaed as he one-year growh rae of he ordinary profi, which is equivalen o income before exraordinary iems and axes. To avoid endogeneiy problems, we employ lagged explanaory variables as insrumenal variables. 15 A panel regression analysis is carried ou according o he resul of he Hausman es, which rejecs he random effec models for boh ime and cross secion, and he fixed ime effec model is hen seleced for equaion (13). 16 11 Among conrol variables, he coefficien of firm size is always posiive and significan. The coefficien of Debasse is significan and negaive due o larger 15 King and Segal (2008) and Akins and Dyl (1997) use lagged explanaory variables as insrumenal variables. The resul of he Hausman (1978) es does no rejec he null hypoheses ha Foreign and Crosshld in he value equaions (10) are exogenous. 16 The Hausman es rejecs he random effec model a p = 0.0099 wih equaion (13) 25
ineres coss. This sudy fails o find any posiive effec of ousider monioring by deb ousanding. Model1 in Table VIII is he resul wihou Foreign, Crosshld, and he inverse Mills raios. Regarding he invesmen horizon s effec on firm value, Horizon is negaively correlaed wih firm value. This finding indicaes ha a relaively shor shareholder invesmen horizon is associaed wih a higher valuaion. We see ha Top30 is posiively correlaed wih firm value. Higher degrees of ownership concenraion are associaed wih higher valuaions. This resul is consisen wih he predicions of Bhide (1993) and Holmsrom and Tirole (1993) abou he behavior of large shareholders srenghening monioring managemen because of difficuly o sell shares in he marke. Model2 and model3 in Table VIII are he resuls including variables for invesor caegories and he inverse Mills raio a he 1 and 10% foreign ownership hresholds, respecively. The effecs of foreign ownership raios and he inverse Mills raios of Foreign are insignifican in model2 and model3. Individual invesors ownership is posiively correlaed wih firm value. The resul is consisen wih Amihud e al. (1999) which find a posiive associaion wih a number of shareholders and sock price. The variable Crosshld is negaively correlaed wih firm value, which means ha ownership concenraion by cross-holders has a negaive impac on firm value, as expeced. If here is a high proporion of cross-holding shareholders, marke paricipans judge he firm o be lax in corporae governance and discoun is 26
value. Absence of corporae governance severely deerioraes a firm s valuaion as well as liquidiy. (Table VIII around here) V. Conclusions We empirically show ha he invesmen horizon of a firm s ownership srucure affecs is liquidiy. In addiion, liquidiy and corporae governance influence he firm s value. Concenraion of ownership measured by he op 30 shareholders ownership raio has a negaive impac on liquidiy. This finding is consisen wih he adverse selecion hypohesis. We find ha he degree of concenraion has a posiive effec on firm value. Our definiion of Top30 is he ownership raio of he op 30 shareholders minus any insider block holdings wihin hem. The variable represens shareholder monioring aciviy o improve firm value. Alhough here is a rade-off beween he coss of illiquidiy and he meris of corporae governance, he resuls sugges ha, in our sample, he posiive effec on firm value brough abou by beer governance exceeds he negaive effec imposed by illiquidiy. The invesmen horizon of a firm iself relaes o is marke liquidiy and value. The firm s laen invesmen horizon is compued from he ownership srucure and he average invesmen horizon of he invesor caegories. The resuls indicae ha he shorer he invesmen horizon, he higher boh he liquidiy and he firm value. This finding is consisen wih he invesmen horizon hypohesis. 27
Invesor caegories, such as foreign, individual, and cross-holding shareholders, have a unique effec on firm liquidiy and value. The greaer he degree of foreign ownership, he higher he liquidiy. This finding is consisen wih prior sudies, which repor ha foreigners invesmen syles are well-diversified (Hochkiss and Srickland 2003) and heir horizon ends o be shor-erm (Tesar and Werner 1995). In he case of cross-holding shareholders, however, we find heir proporions o be negaively relaed wih liquidiy and value. These resuls suppor he invesmen syle and governance hypoheses. The ownership raio of individuals is posiively correlaed wih liquidiy and firm value. This finding suppors he view in which hey are uninformed invesors and heir paricipaion in he marke miigaes adverse selecion coss. Wih respec o firm value, he role of foreigners is somewha vague. Furher invesigaion is necessary in fuure research projecs. 28
References Amihud, Y. 2002. Illiquidiy and sock reurns: Cross-secion and ime-series effecs. Journal of Financial Markes 5, 31-56. Amihud, Y., and Mendelson, H. 1986. Asse pricing and he bid ask spread. Journal of Financial Economics 17, 223-249 Amihud, Y., and Mendelson, H. 2008. Liquidiy, he value of he firm, and corporae finance. Journal of Applied Corporae Finance 20-2 32-45. Amihud, Y., Mendelson, H., and Uno, J. 1999. Number of shareholders and sock price: Evidence from Japan. Journal of Finance 54-3, 1169-1184. Akins, A., and Dyl, E. 1997. Transacions coss and holding periods for common socks. Journal of Finance 52-1, 309-325. Bhide, A. 1993. The hidden coss of sock marke liquidiy. Journal of Financial Economics 34, 31-51. Chordia, T., Roll, R., and Subrahmanyam, A. 2000. Commonaliy in liquidiy, Journal of Financial Economics 56-1, 3-28. Davies, J. R., Hillier, D., and McColgan, P. 2005. Ownership srucure, managerial behavior and corporae value. Journal of Corporae Finance 11, 645-660. Dahlquis, M., and Robersson, G. 2001. Direc foreign ownership, insiuional invesors, and firm characerisics. Journal of Financial Economics 59 413-440. Glosen, L., and Harris, L., 1988. Esimaing he componens of he bid/ask spread, Journal of Financial Economics 21-1, 123-142. 29
Glosen, L., and Milgrom, P., 1985. Bid, ask and ransacion prices in a specialis marke wih heerogeneously informed raders, Journal of Financial Economics 14-1, 71-100. Hausman, L. 1982. Specificaion ess in economerics. Economerica 46-6, 909-938. Holmsrom, B., and Tirole, J. 1993. Marke liquidiy and performance monioring. Journal of Poliical Economy 101-4, 678-709. Hochkiss, E., and Srickland, D. 2003. Does shareholder composiion maer? Evidence from he marke reacion o corporae earnings announcemens. Journal of Finance 58-4, 1469-1498. Huang, R., Soll, H., 1997. The componens of he bid-ask spread: A general approach, The Review of Financial Sudies 10-4, 995-1034, 1997 Kahn, C., and Winon, A. 1998. Ownership srucure, speculaion, and shareholder inervenion. Journal of Finance 53-1, 99-129. Kang, J., and Sulz, R. M. 1997. Why is here a home bias? An analysis of foreign porfolio equiy ownership in Japan. Journal of Financial Economics 46, 3-28 King, M., and Segal, D. 2008. The long-erm effecs of cross-lising, invesor recogniion, and ownership srucure on valuaion. Review of Financial Sudies 22-6, 2393-2404 Mahani, S., Nashikkar, A., Subrahmanyam, M., Chacko, G., and Mallik, G. 2008. Laen liquidiy: A new measure of liquidiy, wih an applicaion o corporae bonds. Journal of Financial Economics 88-2, 272-298. 30
Maug, E. 1998. Large shareholders as moniors: Is here a rade-off beween liquidiy and conrol? Journal of Finance 53-1, 65-98. Morck, R., Nakamura, M., and Shivdasani, A. 2000. Banks, ownership srucure, and firm value in Japan. Journal of Business, 73-4, 539-567 Nia, K. 2000. Cross-sharing and enerprise managemen, posiive analyses concerning influence of ownership srucure (in Japanese). Securiy Analyss Journal 38-2, 72-93 Rubin, A. 2007. Ownership level, ownership concenraion and liquidiy. Journal of Financial Markes 10, 219-248. Sarin, A., Shasri, K., and Shasri, K., 2000. Ownership srucure and sock marke liquidiy. Working paper, Universiy of Pisburgh, Pisburgh, Pennsylvania. Tesar, L., and Werner, I. M. 1995. Home bias and high urnover. Journal of Inernaional Money and Finance, 14-4 467-492. Woodridge, J., 2002. Economeric Analysis of Cross Secion and Panel Daa. MIT Press, Cambridge, Massachuses. 31
Figure 1. Ownership raios of foreign and cross-holding shareholders. 30% 25% 20% 15% Foreigner Anei Mochiai 10% 5% 0% F2002 F2003 F2004 F2005 F2006 F2007 This figure shows he rend in he ownership srucures of Japanese companies for he period 2002 o 2007. The foreigner ownership raio is ha a he end of he fiscal year as repored by Quick s AMSUS. The anei (sabilizing holding) and mochiai raios of each sock are wo kinds of cross-holding raios esimaed by he NLI Research Insiue. A mochiai holding occurs when wo lised companies muually hold shares, as confirmed hrough disclosed maerials. An anei holding refers o he case where he ownership of a firm s shares by banks and life insurance companies is disclosed bu he firm s holdings of counerpary shares canno be confirmed by disclosed maerials. The mochiai raio is used in he res of he paper. The sample includes Firs- and Second-Secion socks of he TSE. 32
Table I. Average invesmen horizon by invesor group. The annual urnover raio by invesor group is he average porfolio marke value a he sar and end of each year divided by he oal rading volume, during he year. Each invesor group s invesmen horizon is an inverse of he invesor group s annual urnover. j j 1 ( porfolio marke value porfolio marke value ) 1 j Invesmen Horizon 2 oal rading volume j (1) Fiscal Non-financial Foreigner Individuals Year Corporaes Trus Banks Insurance Banks 2004 0.414 0.531 7.488 1.238 15.451 11.515 2005 0.331 0.304 5.711 1.065 18.845 13.143 2006 0.303 0.393 6.615 1.206 21.274 17.568 2007 0.219 0.373 6.773 0.959 16.659 15.650 Uni: year. Source: TSE s Share Ownership Survey and Invesmen Trends by Invesor Group. 33
Table II. Summary saisics. This able shows he summary saisics of he explanaory variables. Here, Horizon is a weighed average of he invesmen horizons of sample firms, QRaio is Tobin s q = (aggregae marke value + ineres-bearing deb)/(oal capial + ineres-bearing deb), where he aggregae marke value is adjused for he cross-holding proporion of ownership. FY Mean Median Maximum Minimum Sd Dev. #Obs. Horizon ALL 4.81 4.89 10.12 0.28 1.34 6,695 (Year) 2004 4.72 4.82 8.00 0.28 1.24 1,672 2005 5.29 5.40 10.12 0.43 1.45 1,680 2006 5.29 5.40 10.10 0.43 1.45 1,680 2007 4.66 4.83 8.63 0.31 1.30 1,686 Top30 ALL 50.84 50.60 91.38 0.73 15.74 6,604 (%) 2004 51.16 51.11 91.20 4.96 15.64 1,633 2005 51.40 51.08 91.38 3.89 15.71 1,649 2006 50.46 50.10 90.19 2.59 15.66 1,658 2007 50.36 50.05 88.51 0.73 15.92 1,664 QRaio ALL 1.25 1.05 18.91-9.33 0.88 6,685 (Adjused 2004 1.22 1.05 12.89 0.22 0.77 1,657 Tobin's q) 2005 1.48 1.22 18.91-9.33 1.14 1,667 2006 1.29 1.10 11.90-7.50 0.87 1,676 2007 1.01 0.89 9.26 0.24 0.59 1,685 Crosshld ALL 9.22 7.34 55.22 0.00 8.55 6,695 (%) 2004 9.14 7.40 49.64 0.00 8.32 1,657 Cross- 2005 9.12 7.35 53.02 0.00 8.42 1,672 Holdings 2006 9.25 7.32 55.22 0.00 8.63 1,680 2007 9.36 7.30 53.11 0.00 8.83 1,686 Foreign ALL 12.80 9.48 78.27 0.00 11.91 6,695 (%) 2004 12.63 9.17 73.65 0.05 11.81 1,657 2005 12.55 9.22 77.63 0.05 11.68 1,672 2006 13.18 9.97 77.45 0.00 12.07 1,680 2007 12.84 9.42 78.27 0.00 12.07 1,686 Indiv ALL 32.10 29.46 95.98 1.81 17.06 6,695 (%) 2004 32.90 30.57 93.51 2.90 16.76 1,657 2005 31.48 28.58 95.40 2.02 16.74 1,672 2006 31.71 29.17 95.98 1.81 17.08 1,680 2007 32.32 29.57 95.37 2.85 17.60 1,686 Noe: Some of he minimum values of Tobin s q (QRaio) are negaive due o he negaive equiy capial of he disressed firm. 34
Table III. Correlaions. Panel A. Correlaion beween horizon, size, and urnover. This able shows he correlaion beween Horizon, Top30, Log marke value, and Turnover. Here, Horizon is a weighed average of he invesmen horizons of each firm (equaion(2)) and Marke value is he number of shares ousanding muliplied by he sock price a he end of he fiscal year. Turnover is calculaed as he daily rading volume divided by he number of shares ousanding. The correlaion coefficiens are esimaed annually and averaged over he years 2004 o 2007. Horizon Top30 Log Size Turnover Horizon 1 0.4631 0.2917-0.1108 Top30 0.4631 1 0.0130-0.1347 Log Size 0.2917 0.0130 1 0.0350 Turnover -0.1108-0.1347 0.0350 1.0000 Panel B. Correlaion beween Horizon and holding raios. This able shows he correlaion beween he Horizon and he holding raios of invesor caegories such as foreign, cross-holding, and individual invesors. The correlaion coefficiens are esimaed annually and averaged over he years 2004 o 2007. Horizon Foreign Crosshld Indiv Horizon 1-0.0298 0.3493-0.7557 Foreign -0.0298 1-0.1221-0.4885 Crosshld 0.3493-0.1221 1-0.1224 Indiv -0.7557-0.4885-0.1224 1 35
Table IV. Panel leas squares analysis of illiquidiy. This able shows he relaion of illiquidiy wih marke value, invesmen horizon, ownership concenraion, and he ownership raios of foreigners and cross-holding for Firs- and Second-Secion socks on he TSE over he period 2004 o 2007. The resuls are from he following panel leas square regressions. The fixed period effec model is seleced as a resul of he Hausman es. The basic model specificaion is (8) RILLIQ Crosshld 5 a 1 Horizon 2Top30 6 Log_Size Foreign ADRdummy 7 3 4 Indiv where RILLIQ is he relaive ILLIQ for firm j defined by equaion (7), Log_Size, is he naural logarihm of firm j s marke value a he end of March, Horizon is he weighed average of he holding period for firm j s sockholders in year, and Crosshld and Foreigner are he percenages of ownership of firm j s invesor caegories a he end of fiscal year. The American Deposiary Receip (ADR) dummy equals 1 if a firm issues ADRs, and 0 oherwise. Here, is an error erm. Heeroskedasiciy is correced by Whie diagonal sandard errors and covariance correcions. Dependen Variable: R_ILLIQ Year Model1 Model2 Model3 Model4 Coefficien -value Coefficien -value Coefficien -value Coefficien -value Horizon 0.1575 13.00 0.0662 4.47 TOP30 0.2287 2.64 0.0059 6.43 Horizon^2 0.0173 14.82 0.0082 5.66 TOP30^2 0.2797 3.36 0.5745 6.52 Foreign -0.0259-10.45-0.0254-10.40 CrossHld 0.0113 6.70 0.0114 6.74 Indiv -0.0173-6.17-0.0178-6.47 Log Size -1.0620-128.56-1.0633-130.27-0.9828-74.00-0.9904-73.55 ADRdummy -0.2969-8.58-0.2859-8.20-0.2807-8.39-0.2711-8.08 Inercep 2.4488 37.08 2.8237 59.28 2.6161 34.56 2.8902 46.18 Adjused R-squared 0.814 0.803 0.820 0.819 Observaions 4,899 4,899 4,899 4,899 36
Table V. The probi selecion model The probi selecion model for foreign and cross-holding shareholders are esimaed. The explanaory variables are a firm s marke value(log_size), is overseas sales(expor),reurn on equiy (ROE), and he ADR dummy, a dummy variable ha equals 1 if he firm issues ADRs and 0 oherwise. The dependen variable in model1 (model2) is 0 when foreigners, as a group, own less han 1% (10%) of he firm, and 1 when hey hold 1% (10%) or more. The dependen variable in model3 is 0 when here are no cross-holding shareholders, and 1 when hey hold more han 0% of shares. The cross-holding model esimaions (model3) include he number of shares no held by managemen or cross-holders ( Floa )as an explanaory variable. The selecion models for foreign ownership 1%(10%) are (9) Forein _1%(10%) 4 ROE Growh 5 a Log_Size 1 Horizon 2 Expor 3 ADRdummy The selecion model for crossholding is (10) Crosshld _ dummy Growh 4 a Log_Size Floa 5 1 Horizon 2 Expor 3 ADRdummy Model1 Model2 Model3 Dependen Variable Foreign>= 1% Foreign>=10% Crosshldf>0% Coefficien z-saisic Coefficien z-saisic Coefficien z-saisic Log_Size 0.7606 25.75 0.5737 37.97 0.1446 8.27 Expor 0.8842 4.84 1.3799 14.75 0.4683 4.39 ADRdummy 0.0058 0.02 0.2121 2.06-0.2514-2.68 ROE 0.0012 1.19 0.0056 4.26 Growh -0.0442-1.62-0.0810-3.27-0.0246-0.98 Floa 0.0139 6.64 Inercep -1.0712-13.29-2.51137-41.59 0.3704 3.95 Adjused R-squared 0.5123 0.3140 0.0249 Observaions 6,695 6,695 6,695 37