R Square Measure of Stock Synchronicity

Transcription

1 Internatonal Revew of Busness Research Papers Vol. 7. No. 1. January Pp R Square Measure of Stock Synchroncty Sarod Khandaker* Stock market synchroncty s a new area of research for fnance and economcs lterature. Morck at al (2000) are among the frst to propose a model of stock market synchronous behavour of emergng markets and suggested that R square values of emergng markets are hgher than the developed economes. Ths study analyses stock market data of eght emergng economes and three developed natons, and fnds evdence that R square values of emergng economes are hgher than ther counterparts. It s also found that hgher nflaton, country geographcal sze, low level of corporate governance mechansm and nflaton causes stock prce to move n the same drecton n observed emergng economes. Keywords: Synchroncty, R square, Correlaton, panel data 1. Introducton Emergng markets exhbt hgher stock prce synchroncty than the developed economcs. Stock synchroncty refers the tendency of a stock market to move n the same drecton n a partcular perod of tme; such as a gven day or week. However, the stock synchroncty could be upward or downward dependng upon the overall movement of the stock market. Morck et al.(2000) are among the frst who propose two models to measure stock return synchroncty over a partcular perod of tme. Both the models capture the level of frm specfc nformaton that s reflected n ndvdual frm share prces. In addton, Khandaker and Heaney (2008) use a large tme seres data and replcate both these model to measure the stock market synchroncty. They suggested that hgher nflaton, country wde corrupton and poor corporate governance mechansm causes hgher stock market synchroncty n emergng economes. Ths paper partcularly uses the second model of stock market synchroncty suggested by Morck et al.(2000) and Khandaker and Heaney (2008) to analyse the selected emergng economes. The study uses eght emergng countres and three developed economcs for the analyss. It s found that country specfc characterstcs such as voce & accountablty, corrupton ndex and geographcal sze are negatvely correlated wth R square measure and nflaton s postvely correlated consstent wth Morck et al.(2000). In addton, Chna, Malaysa and Turkey exhbt the hghest R square synchroncty durng the study perod, where as Germany and Japan exhbt the lower level of R square values. Fnally there s evdence that hgher R square values are evdent n emergng markets. *Dr Sarod Khandaker, Faculty of Busness and Enterprse, Swnburne Unversty of Technology, E- mal address: skhandaker@swn.edu.au

2 2. Recent Lterature Morck et al. (2000) are among the frst who successfully argue that share prces n emergng economes move more closely together than the developed markets. They use b-weekly stock return data to measure stock return synchroncty for 40 countres around the world. They argue that a lack of corporate transparency, less nformaton dsclosure polcy, corrupton and less corporate governance mechansms nfluence the poorer economes stock market to move n the same drecton. They found that poorer economes often do not have well structured fnancal systems and less dsclosure polcy mght nfluence the stock market to follow the overall market n a partcular drecton n a gven week. Further, Morck et al. (2000) found that stock prce synchroncty s negatvely correlated wth a country s sze. They suggest that a small country effect mght cause the hgher stock prce co-movement found n smaller countres. Ths fndng s further supported by Levne and Zervos (1998) and Khandaker and Heaney (2008) who also argue that smaller countres often have under developed fnancal markets leadng to lower fnancal growth. However, ths s to note that Khandaker and Heaney (2008) use a longer tme perod and large set of stock market data to analyse the stock market synchronous behavour, where as Morck et al. (2000) use a large set of countres over a short tme perod of tme. There s also evdence that ndvdual stocks are becomng more synchronous for average countres. For example, L et al. (2003) examne the Canadan stock market and compare ths wth Mexco and the East Asan markets. L et al. (2003) found that Canadan stocks move less synchronously than Mexcan stocks. They argue that snce Canada entered nto the free trade agreement wth the US, ts stocks have exhbted a permanent ncrease n frm-specfc varaton because of ncreased market openness to the US market. In addton, Skafe et al. (2006) fnd that lower R-square values are assocated wth more nformatve prces n developed economes, where as hgher R square values are assocated wth low nformatve prces n the stock market and lower corporate governance mechansm ( Nguyen and Aman 2006, Wurgler 2000). It s to note that the concept of stock market synchroncty s not very new to the fnance lterature. For example, French and Roll (1986) and Roll (1988) argue that hgh stock prces and a well-nformed market generate low stock synchroncty. Roll (1988) shows that the movement of stock prces depends on several related factors whch nclude frm-level and market-level nformaton that s captalsed nto stock prces. However, t s found from the exstng lterature that stock markets n emergng economes are more synchronous over tme than n developed markets. Hgher stock synchroncty s evdent from countres that do not respect prvate property rghts, and corporate governance mechansms are less effectve n those economes. 166

3 3. Research Methodology and Data 3.1 Research Methodology The study uses synchroncty measure developed by Morck et al.(2000) called the R square measure. The R-square synchroncty measure s the most popular model n the lterature for capturng stock market synchroncty. For Example, Roll (1988) found that ndvdual stocks n the USA exhbt low R-square statstcs due to the hgher dsclosure polcy. Chan and Hameed (2006) also use ths measure to explan analyst coverage for emergng markets and argue that emergng economes exhbt hgher stock prce co-movement due to the hgh cost of collectng frm-specfc nformaton. Further, Morck et al. (2000) argue that lower frm-specfc nformaton s produced n emergng markets whch results n hgher R-square values, whle hgher frmspecfc nformaton produced n developed markets results n lower R-square values. In addton, Jn and Mayers (2004) argue that countres wth hgher R- square values experence more frequent market crashes. However, gven the smple market model share return can be expressed as a fracton of share market return: R, t Rm, t, t (1) where t perod, R, t s the frm return for perod t, R m, t, t s the error term and 1 and s the market return of frm for are estmated parameters. The 2 R measure s the percentage of varaton n weekly return of stock n country j explaned by varatons n country j s market return, or: R Cov ( R ) 2 Rm t m 2 (2) where Cov( R returns and Rm ) s the covarance between the share returns and share market s the standard devaton for asset. A hgh R-square ndcates a hgh degree of stock return synchroncty and a low R-square ndcates a low degree of stock return synchroncty for a gven stock f or a partcular perod of tme, e.g. a gven week or day Data The study analyses stock market data from eleven economes, ncludng three developed economes and eght emergng markets. The data span the perod from January 1996 to December It s to note that the study uses DataStream database to collect weekly stock return data. However the study also uses yahoo 167

4 fnance database where a gap exsts n DataStream database. The study uses both the lve stocks and the dead stocks to mnmze the survvorshp bas problem. Table 1 llustrate the total number of observed frm data collected from the DataStream and yahoo fnance database. In total the study analyses 6.6 mllon weekly frm observatons for estmatng R square measure for developed and emergng countres. Table 1 Total number of observed frms for R square synchroncty analyss Emergng Country No of Frms Developed Country No of Frms Argentna 216 Australa 2,709 Chna 1,905 Germany 1,506 Cyprus 144 Japan 2,842 Malaysa 1,316 Sr Lanka 301 Tawan 1,145 Turkey 515 Zmbabwe 100 Total number of frms 12, Panel Data Analyss Varables The study uses sx country-level stock synchroncty explanatory varables for panel data analyss; these nclude voce and accountablty, regulatory control, GDP per capta, nflaton, geographcal sze of a country and corrupton ndex. Corporate governance ndces e.g. voce & accountablty and regulatory control collected from the World Bank corporate governance database. These governance ndcators reflect the statstcal complaton of responses to questons concernng the qualty of governance. It s to note that, the World Bank uses sx corporate governance ndcators for measurng the good governance system, whch s dvded nto three clusters. The governance ndcators are measured n unts rangng from -2.5 to 2.5. Hgher values correspond to better governance outcomes and lower values correspond to poorer outcome. For example, a country that ranks 2.0 n terms of corporate transparency mantans strong transparency n government and prvate sectors. Ths also ndcates that there s a strong flow of nformaton n the market. In contrast, f a country has a rank of -2.0 for the rule of law ndces, ths would ndcate a relatvely poor qualty legal system. The study also collected GDP per capta and nflaton data from World Bank database. Further, corrupton ndexes data are collected from Transparency Internatonal database a German based organzaton. Transparency Internatonal frst presented ther corrupton ndex (CP) n The corrupton ndex s based on the past three year s corrupton percepton data. The corrupton ndex uses publc opnon surveys to measure the level of corrupton n a country and the corrupton scale ranges from 1 to 10 ponts. Countres wth greater transparency are awarded hgher ponts (maxmum 10) and countres wth lower transparency are awarded fewer ponts (mnmum 0). 168

5 In addton, geographcal sze of a country data s collected from CIA world factbook webste. The CIA world factbook s publshed by the Central Intellgence Agency, USA. The CIA webste provdes accurate and relable country level data ncludng geographcal sze. 4. Descrptve Statstcs Analyss Table 2 llustrates the descrptve statstcs of R-square measure for eght emergng countres. The data span the perod from January 1996 to December It s found that Chna, Malaysa, Turkey and Tawan exhbt hgher R square measure stock synchroncty for the emergng country group. Chna exhbts.241 and Malaysa exhbts.254 R-square values durng the observaton perods whch are the hghest R-square synchroncty durng the observaton perod for any emergng countes wth standard devaton and Turkey on the other hand exhbts R-square values wth standard devaton of Ths suggests that Turkey, Malaysa and Chna have the hghest R-square value durng the observaton perod, a result whch s consstent wth prevous lteratures. Table 2 Descrptve Statstcs for Emergng Country Group Mean Medan Maxmum Mnmum Std. Dev. Argentna Chna Cyprus Malaysa Sr Lanka Tawan Turkey Zmbabwe Average Further, Chna and Tawan exhbt R-square values of maxmum durng the observaton perod whch s unexpected. Ths suggests R-square measure of stock synchroncty could be as hgh as 100 percent for a gven week n emergng countres. Ths s a surprsng fndng though consstent wth Morck et al (2000). Table 3 llustrate R-square measure of observed developed countres. It s found that Germany and Japan exhbt lower level of R-square value durng the observaton perod. However, the average R-square synchroncty measure for the developed country group s 0.019, qute lower than the emergng country group (0.177). Ths suggests that the observed developed country group exhbt lower level of stock synchroncty than the observed emergng country group durng the study perod. 169

6 Table 3 Descrptve Statstcs for Developed Country Group Mean Medan Maxmum Mnmum Std. Dev. Australa Japan Germany Average The study also run ANOVA test statstcs to examne whether there are any statstcal mean dfferences n the R-square measure between the developed country and the emergng country group. Table 4 llustrates ANOVA test statstcs between two groups. It s found that there s a statstcal mean dfference n the R- square synchroncty measure at one percent sgnfcant level between two groups. Table 4 ANOVA test Statstcs for the R-square Measure. Sum of Squares Degrees of Freedom Mean Square F- Statstcs Sg. Between.054 Groups Wthn Groups Total Sub-Perod Data To check the possblty of changes n the level of R-square synchroncty measure durng the observaton perod from 1996 to 2005, the study dvde the tme seres nto fve sub-perods, , , , and Table 5 llustrates the descrptve statstcs of R-square synchroncty measure over the fve sub-perods for the emergng county group. Table 5 Sub perod R-square synchroncty measure for emergng countres Form January 1999 to 2005 Country R-square R-square R-square R-square R-square Argentna Chna Cyprus Malaysa Sr-Lanka Tawan Turkey Zmbabwe Average Studes fnd that countres exhbt hgher R-square measure durng the full perod, also exhbt the hgher synchroncty for the sub perods. Examples nclude Chna 170

7 and Turkey. However, the levels of R-square measure stock synchroncty reman volatle n sub perods for most of the emergng countres. Table 6 llustrates R-square varaton by sub-perods for the developed country group. It s found that developed countres exhbt lower level R-square values durng the observed sub-perods, especally Japan exhbts lowest R-square values n all sub perods. It s to note that the result s conflctng for Japan wth the prevous studes. Table 6 Sub Perod R square Synchroncty for developed Countres Form January 1999 to 2005 Country R-square R-square R-square R-square R-square Australa Germany Japan Average However, Japan has a hstory of volatle stock market and R-square stock synchroncty measure only the stock market co-movement rather than stock market volatlty. It s possble for a volatle stock market to exhbt low level of stock market synchroncty, e.g. Japan. Further analyss of ths queston remans for future research. 5. Panel Data Analyss The model The study uses ordnary least square model and whte adjusted standard errors n a fxed effect model for the panel analyss. Table 7 llustrates cross sectonal panel analyss usng both models. The study also tests correlatons between the explanatory varables to check the multcollnearty between the explanatory varables. However, the paper used the followng model to explan R-square measure of stock synchroncty. RSqu RC VC IN CP GDP log( SIZE ) (3) Where RSqu represents R-square measure for stock synchroncty of country and α s a constant. RC s the regulatory control ndex, accountablty ndex, IN s the nflaton, gross domestc product per capta / 10,000, CP s the corrupton ndex, VC s the voce and GDP s the s SIZE s geographcal sze and the error term. The natural log of geographcal sze s used to lmt the effect of skewness on ths varable. 171

8 Table 7 Panel data analyss for R square synchroncty measure Here, the frst column ncludes varable names, second column ncludes pooled regresson fndngs and thrd column ncludes fndngs of the fxed effect model. Two values are reported below the estmated coeffcent. The frst n parenthess s the t statstc usng whte adjusted standard errors and the second, n the brackets, s the P value for the statstc. Varables Pooled regresson model Fxed effect model Regulatory Control (3.987) (3.161) [0.000] [0.003] Corrupton Index (-1.573) (-1.604) [0.124] [0.117] GDP Per Capta / (-2.706) (-3.151) [0.010] [0.003] Inflaton (0.782) (0.671) [0.439] [0.507] Voce & Accountablty (-2.947) (-5.033) [0.005] [0.000] Log (Geographcal Sze) (-0.867) (-1.233) [0.391] [0.225] R-square The study fnds evdence that corrupton ndex s negatvely correlated wth the R- square synchroncty measure at the 15 percent sgnfcance level usng the pooled regresson analyss. It s a sgnfcant result for the study because t s argued that hgher corrupton rate causes hgher stock prce synchroncty thus hgher R square. Further, t s found that emergng countres rank lower n corrupton ndex and developed economes rank hgher due to the level of corrupton. The fndngs are consstent wth Morck et al. (2000) and Khandaker and Heaney (2008). In addton, Khandaker and Heaney (2008) argued that hgher corrupton ndcates weaker poltcal system and cvl rghts whch causes corrupt economes to rank lower n the property rghts ndex. Regulatory control s an mportant governance mechansm thought necessary to establsh a good governance system n a country. The study fnds that regulatory control s postvely correlated wth R-square measure at the one percent 172

9 sgnfcance level. It s to note that both the governance mechansms are mportant to establsh a good governance system n a country. The fndng s sgnfcant as t s argued that countres wth good governance mechansm have structured captal markets thus exhbt lower stock prce synchroncty. Another governance mechansm voce and accountablty s negatvely correlated wth the R-square measure at one percent level whch s surprsng. Nevertheless, the study fnds evdence that some of the observed emergng economes, such as Tawan and Cyprus ranked hgher n voce and accountablty ndex than some of the developed natons (e.g. Japan) though exhbt hgher R-square values durng the observaton perod. Conversely, the study also uses whte adjusted standard error n a fxed effect model for the panel analyss. The results are unchanged wth small varaton n coeffcent. Unexpectedly GDP per capta s postvely correlated wth the R-square measure at the one percent sgnfcant level usng both the models. Ths s a surprsng result as t s assumed that hgh GDP per capta countres produce lower level of stock market synchroncty and lower R-square values. However, the coeffcent for GDP per capta s close to zero suggestng mmateral mpact to ths model. Geographcal sze s negatvely correlated wth the R-square synchroncty measures and nflaton s postvely correlated. Ths result s consstent wth Khandaker and Heaney (2008) and Morck et al. (2000) who suggested that country geographcal sze s generally negatvely correlated wth stock market synchroncty. However, the mpact of both these varables on R-square measure synchroncty s not statstcally sgnfcant and co-effcent of both these varables are close to zero. 6. Dscusson It s found from the analyss that the R-square values for the developed economes are less than the emergng economes usng the sample countres. In total, the study uses 6.6 mllon weekly frm observatons and 12,699 frm data from eght emergng economes and three developed markets. The study fnds evdence that emergng economes exhbt R-square values durng the observaton perod compared to the R-square values of from the developed economes. There s evdence that emergng economes lke Chna and Malaysa could have 100 percent of the stock to move n the same drecton n a gven week thus havng R-square values of t s a very sgnfcant result and consstent wth the fndngs of Morck et al. (2000), Alper and Ylmaz (2004) and Khandaker and Heaney (2008). In addton, Japanese stock market exhbts lowest R-square values durng the observed full perod and sub perod analyss suggestng that stock market volatlty and synchroncty are not to capture smlar aspects of stock market behavour. It s found that R-square values are lower n hgh nflatonary economes than low nflatonary economes though the result s not statstcally sgnfcant. Morck et al. (2000) argued that hgh nflaton causes hgh stock prce synchroncty n emergng economes. Further, the study fnds evdence that emergng countres lke Turkey exhbts hgher R-square values and also has exhbt hgh level of nflaton. 173

10 However, n recent years Turkey exhbts lower level of nflaton whch also affects ts R-square values n sub perod and Corporate governance mechansms have strong mpact on R-square synchroncty values. The study uses two corporate governance mechansms and found that voce and accountablty s sgnfcantly assocated wth the R-square measure values. Prevous studes also show that countres wth hgh corporate governance mechansm exhbt a low level of stock market synchroncty. Examples nclude Germany and Japan. In addton, countres do not respect the prvate property rghts has hgh level of corrupton and also ranked low n corporate governanc e ndex, thus exhbts a hgh R-square values. Example ncludes Chna and Turkey (La-Porta et al. 1998). There s also evdence that country geographcal sze has some mpact on ts R- square values thus stock synchroncty measure. For example, t s found that most of the countres of the world conduct the major part of ther tradng nternally and large countres often have structured debt and equty markets whch drectly nfluence the growth of the economy; such as Australa and the USA. In contrast, smaller countres tend to have fewer large frms n ther captal market whch can manpulate the fnancal markets and could lead to hgher stock prce synchroncty (Morck et al 2000). However, t s assumed that the geographcal sze mpact s not statstcally sgnfcant n ths case due to the small number of large country n the sample. The study presented evdence that stock market synchroncty s hgher n emergng markets than the developed economes usng the R-square measure. There s also evdence that hgh nflaton, country geographcal sze, low level of corporate governance mechansm and nflaton causes stock prce to move n the same drecton n emergng economes. Although ths study uses a large tme seres data from eght emergng economes and three developed markets, a further research s necessary to dentfy the country specfc characterstcs of emergng market synchronous behavour. Endnotes L et al. (2003) use R-square synchroncty measures for ther cross-sectonal analyss. References CIA. 2007, "CIA World Factbook." Washngton D.C. Durnev, A, Morck, R, Yeung, B and Zarown, P 2003, "Does Greater frm-specfc Return Varaton Mean More or Less Informed Stock Prcng?" Journal of Accountng Research., vol. 41, no. 5, pp French, KR and Roll, R 1986, "Stock Return Varances: the arrval of nformaton and the reacton of traders." Journal of Fnancal Economcs., vol. 17, no. 1, pp Jn, L and Mayers, S 2004, "R Square Around the World: New Theory and New Tests." Journal of Fnancal Economcs., vol. 58, no. 1, pp

11 Karoly, GA and Stulz, RM 1995, "Why do Markets Move Together? An Investgaton of U.S.-Japan Stock Return Comovements usng ADRS." Journal of Busness and Economc Statstcs., vol. 11, no. 2, pp La-Porta, P, Lopez, FS, Andre, S and Robert, VW 1998, "Law and Fnance." The Journal of Poltcal Economy., vol. 106, no. 6, pp La-Porta, P, Lopez, SF, Shlefer, A and Vshny, R 1997, "Legal determnants of external fnance." Journal of Fnance., vol. 52, no. 3, pp Levne, R and Zervos, S 1998, "Stock Markets, Banks, and Economc growth." Amercan Economc Revew., vol. 88, no. 3, pp L, K, Morck, R, Yang, F and Yeung, B 2003, "Tme varyng synchroncty n ndvdual stock returns: a cross-country evdence." Unversty of Alberta Busness School. Longn, F and Solnk, B 1995, "Is the correlaton n nternatonal equty returns constant: ?" Journal of Internatonal Money and Fnance., vol. 14, no. 1, pp Morck, R, Yeung, B and Yu, W 2000, "The nformaton content of stock markets: why do emergng markets have synchronous stock prce movements?" Journal of Fnancal Economcs., vol. 58, no. 1, pp Nguyen, P and Aman, H 2006, "Corporate Governance and Stock Performance of Japanese Frms." Socal Scence Research Network. Roll, R 1988, "R²", Journal of Fnance., vol. 43, pp Skafe, AH, Gassen, J and La.Ford, R 2005, "Does stock prce synchroncty represent frm-specfc nformaton? The nternatonal evdence." MIT Sloan Workng Paper., Wurgler, J 2000, "Fnancal Markets and the allocaton of Captal." Journal of Fnancal Economcs., vol. 58, no. 1-2, pp