Internatonal Research Journal of Fnance and Economcs ISSN 1450-2887 Issue 6 (2006) EuroJournals Publshng, Inc. 2006 http://www.eurojournals.com/fnance.htm The Stock Market Reacton to Extreme Events: The Evdence from Turkey Huseyn Aktas Celal Bayar Unversty, Turkey The Faculty of Economcs and Admnstratve Scences Department of Accountng and Fnance, Uncubozkoy/Mansa E-mal: huseyn.aktas@bayar.edu.tr Tel. (90) (0512)542-4225482, Fax, (90) (236) 233-2729 Semra Oncu Celal Bayar Unversty, Turkey The Faculty of Economcs and Admnstratve Scences Department of Accountng and Fnance, Uncubozkoy/Mansa E-mal: semra.oncu@bayar.edu.tr Abstract Effcent Market Hypothess (EMH) assumes that market prces reflect all avalable nformaton and expectaton, and that any new nformaton s properly ncorporated nto prces wthout any delay. Sharp fall of securty prces durng the turbulent tmes s a very unque opportunty for testng valdty of EMH. Unexpected events can put more stress on the fnancal market, and market partcpants may lose ther ablty to asses ratonally the valuaton mplcatons of event. On March 1, 2003 Turksh parlament rejected the hghly controversal bll that allows the deployment of U.S Troops n Turkey. Ths paper nvestgates the prcng behavor of the Turksh stock market n case of a major poltcal event that had strong economc mplcaton for market partcpants. In response to unfavorable poltcal events, stock prces are expected to behave dfferently n the effcent market snce new nformaton wll have dfferent economc mpact on ndvdual frms. The emprcal results presented n ths paper generally support the assumptons of EMH. Key words:extreme events, stock market, market effcency JEL Classfcaton: G12, G14 Introducton In fnance, Effcent Market Hypothess (EMH) s wdely tested proposton by researchers. An mplcaton of EMH s that market prces reflect all avalable nformaton and expectatons, and that any new nformaton s properly ncorporated nto prces wthout any delay. A stock market s speedness to ncorporate new nformaton nto prces s referred to nformatonal effcency. Market s ablty to reflect new nformaton properly s referred to market ratonalty. A ratonal market s one n whch stock prces unbased estmaton of fundamental values of fnancal assets. After the ntroducton of EMH nto the fnancal economcs lterature, nformatonal effcency of stock market has been nvestgated extensvely n volumnous studes. Recently the valdty of market effcency s serously crtczed n the behavoral fnance lterature. The behavoral fnance
79 Internatonal Research Journal of Fnance and Economcs - Issue I (2006) lterature s bascally dvded nto two prmary areas: The dentfcaton of anomales n the EMH and the dentfcaton of ndvdual nvestor s bases whch are not consstent wth classcal theores of ratonal behavor (Pompan and Longo, 2004). The one stream of research related to market effcency concentrates on extreme and dramatc events. The examples of these events are stock market crashes, poltcal events, earthquakes, terrorst attacks etc. On March 1, 2003 Turksh parlament rejected the hghly controversal bll that allows the deployment of U.S Troops n Turkey. The surprsngly rejecton of the bll shook the poltcal and economc relatons between Turkey and USA. Frst tradng day followng the refusal of the bll, The Istanbul Stock Exchange s (ISE) man ndexes went down, nterest rates started to rse, and the roll over of government s debt was n danger. The purpose of ths paper s to examne the prce dscovery process of the Turksh stock market after the rejecton of bll authorzng the deployment of U.S Troops n Turkey. The remander of ths paper s organzed as follow. Secton II revews the lterature related to fnancal market s reacton to extreme and unexpected events. Secton III provdes a bref descrpton of event and presents sample data. Secton IV presents methodology and emprcal results and s followed by concluson. 2. Lterature Revew Sharp fall of securty prces durng the turbulent tmes s a very unque opportunty for testng EMH. Unexpected events can create more stress n the market, and market partcpants may lose ther ablty to asses ratonally the valuaton mplcatons of event. In ths drecton, a number of papers have examned the securty prce behavor when uncertanty n fnancal market ncreases after dramatc events. 2.1. Behavor of Securty Prces around the Market Crash Lmmack and Ward (1990) rase an nterestng queston regardng the usefulness of fnancal models both n explanng and n predctng share prce behavor n the 1987 crash perod. Usng a sample of 270 U.K stocks, they nvestgate the explanatory power of a number of competng models, ncludng the market model and a factor based models. The results of the test of the market model suggest that systematc rsk had a sgnfcant nfluence on the U.K stock s movement, especally n October 1987 n whch a major prce adjustment occurred. Ther fndngs were strengthened when the test were conducted on portfolo data. Usng the sample of Standard & Poor s 500 stocks, Harrs and Spvey (1990) examned the relatve reacton of ndvdual stocks to the market crash on October 19, 1987. They found a sgnfcant relaton between the percentage declne n stock prces and the stocks hstorcal betas. Because the declne n stock prces was less than the declne mpled by the stock s betas, they concluded that behavor of stock market appears between effcency and rratonalty. Fenberg and Tokc (2002) studed two extreme sngle-day declnes n stock prces due to the Asan crss on September 1, 1998 and on September 11, 2001. Usng the 30-stock DJIA data, they showed that on both dates, stocks wth hgher betas decreased relatvely more n a sngle day than stocks wth lower betas. Smlarly, stocks wth hgher betas were found to ncrease relatvely more n a sngle-rse n the stock market than stocks wth lower betas. Kryzaowsk, Swtzer, and Jang (1995) nvestgated the abnormal return, volatlty, and resdual rsk premum behavor of portfolos durng the Canadan stock market crash of 1987. One of ther fndngs s that the performance of beta-sorted portfolos over varous tme ntervals around the crash s nversely related to systematc rsk. Based on the fact that Real Estate Trusts (REITs) are low rsk/low return nvestment vehcles, Glascock, Mchayluk, and Neuhauser (2004) tested the rskness of REITs durng the stock market
Internatonal Research Journal of Fnance and Economcs - Issue 6 (2006) 80 declne of October 1997. They found that the declne n REITs stock values s about one-half as large as the declne of non-reit stock. Ther fndngs about REIT stocks are consstent wth EMH. 2.2. Reacton of Securty Prces to Dsasters Hll and Schneewes (1983) examned the effect on the stock return of publc utlty frms of Three Mle Island nuclear accdent. Ther study ndcates that the mpact of the accdent on non-nuclear frms was less than that on nuclear based utltes. Kalra, Henderson, and Ranes (1993) nvestgated the market reacton n U.S. to the Chernobyl nuclear accdent. Accordng to ths study, utlty nvestors correctly nterpreted that Chernobyl marked the start of a new era for nuclear power n U.S. and the stock market quckly recognzed the consequences for utlty stocks. Maloney and Mulheren (2003) provded evdence on the speed and accuracy of prce dscovery by studyng stock return and tradng volume surroundng the crash of the space shuttle Challenger. They showed that prce dscovery occurred wthout large tradng profts and much of prce dscovery occurred durng a tradng halt of the frm responsble for faulty component. Another research (Blose et al., 1996) conducted on the Challenger crash showed that on the day of exploson, there was a sgnfcant negatve abnormal return on the stocks of NASA contractors, and any contagon effect was lmted to narrow set of NASA contractors. Shelor, Anderson, and Cross (1990) examned the effect of Calforna earthquake on the stock value of frms n the real estate ndustry. The fndngs showed that the earthquake transmts mportant new nformaton to the market that was reflected n statstcally sgnfcant negatve stock returns among those frms operatng n the area ht by the earthquake. Real estate-related frms operatng n other areas were generally unaffected by the earthquake. Angbazo (1996) examned the mpact of Hurrcane Andrew and a subsequent change n the regulatory envronment on the stock prces of 48 publcly-traded property-lablty nsurers. He found that Andrew and the related regulatons had ndustry-wde contagon effect snce they sgnfcantly affected most nsurers, regardless of whether these frms had any loss exposure n the hurrcaneaffected places. Also Lamb (1998) examned the market effcency around hurrcanes Andrew and Hugo. Hs observaton ndcates that the stock market showed an ablty to dscrmnate by the magntude of hurrcane and by Property and Casualty frms based on ther degree of loss exposure. Carter and Smkns (2004) nvestgated the reacton of arlne stock prces after the September 11 attacks. Ther study supports the hypothess of ratonal prcng and suggests that the stock market dfferentated among varous arlne frms. 2.3. The Relaton between Poltcal Rsk and Securty Prces Le and Zak (2006) constructed a portfolo choce model that relates captal flght to return dfferentals, rsk averson, and three types of rsk: economc rsk, poltcal nstablty, and polcy varablty. Estmatng the equlbrum captal flght equaton for a panel of 45 developng countres over 16 years, ther fndngs ndcated that all three types of rsk have a statstcally sgnfcant mpact on captal flght. They also dentfed that poltcal nstablty s the most mportant factor assocated wth captal flght, and several poltcal factors reduce captal flght ostensbly by sgnalng that market-orented reforms are mmnent. Usng analyst estmates of poltcal rsk, Damonte, Lew, and Stevens (1996) concluded that average returns n emergng markets experencng decreased poltcal rsk exceed those of emergng markets experencng ncreased poltcal rsk by approxmately 11%. Furthermore, fndngs showed that the dfference of the mpact of poltcal rsk n emergng and developed markets s statstcally sgnfcant. Blson, Bralsford, and Hooper (2002) presented evdence that poltcal rsk may be able to explan some of the varaton n emergng stock market returns at both country and aggregate portfolo levels. Ther fndngs reveal that poltcal rsk s mportant n explanng return varaton n ndvdual emergng markets, partcularly n the Pacfc Basn, but not n developed markets.
81 Internatonal Research Journal of Fnance and Economcs - Issue I (2006) Dar, Feng, and Chun (2005) nvestgated the possble mpact of poltcal events on Tawan s stock market prces. They found that prce reactons to most of poltcal events are rather nsgnfcant, mplyng those events be largely unnformatve wth only a few exceptons. Abnormal return behavors are also frequently comparable between frms wth small and large foregn nsttutonal ownershp. Mehdan, Nas, and Perry (2005) tested the predcton of two competng theores about stock market s reacton to unexpected poltcal and economc event n Turkey. Ther study showed that Turksh Stock Market systematcally adjust securty prces below ther fundamental values n case of unexpected events. 3. A Bref Descrpton of Event and Data Before the Turksh Parlament narrowly rejected of government s moton, Turkey and U.S reportedly reached an agreement on economc, poltcal, and mltary ssues related to Iraq crses and economc ad to buffer the mpact of a war. At the frst tradng day (March 3, 2003) after unexpected parlament s rejecton, IMKB natonal-100 ndex lost %12.5. It was the fourth worst fall n ts hstory. Only after the government s declaraton that t would mplement an emergency package and possblty of new moton, fnancal market stablzed. To examne whether Turksh Stock Market effcently assessed the valuaton mplcatons of events on March 3, 2003 after the rejecton of bll, Daly Prces for 50 stocks lsted n ISE Natonal-50 ndex as of March 2003 and the ISE Natonal-100 ndex were obtaned from the Fn.net Data Base (www.fnnet.com.tr) over the perod of 20.02.2002-07.03.2003. Stocks n the ISE Natonal-50 ndex were chosen n order to create a sample of stocks for whch market dsplays the greatest breadth and depth. Daly returns are calculated as logarthmc dfferences of daly closng prces. 4. Methodology and Emprcal Results Accordng to EMH, market partcpant should recognze the dfferent mplcatons of new nformaton on stock prces. In terms of systemc negatve news lke a rejecton of moton, market should ncorporate ths new nformaton nto prces by dfferng systematc rsk of each stock. To test the hypothess that percentage drop n ndvdual stock prces should be related to systematc rsk of each stock, we estmated betas of each stock by usng the market model. Based on 60, 120, and 240 daly observatons (leavng the two weeks as a pre-event perod), the market model s used as followng ordnary least square regresson form n estmaton of betas. Rt = α + β Rmt + ε t (1) where: R = Daly returns to stock calculated over perod t t R = Daly returns to the market ndex calculated over perod t mt α = The ntercept β = Systematc rsk of stock ε t = A zero mean random dsturbance term t = 60 days, 120 days, and 240 days Snce heteroscedastcty was detected through applcaton of the ordnary least square regresson, heteroskedastcty-corrected market model s estmated by usng gretl econometrc software. After runnng an OLS regresson, gretl software saves the resduals. The logs of the squares of these resduals then become the dependent varable n an auxlary regresson; ndependent varables of regresson are the orgnal ndependent varables plus ther squares. The ftted values from the auxlary regresson are then used to construct a weght seres, and the orgnal model s re-estmated usng weghted least squares.
Internatonal Research Journal of Fnance and Economcs - Issue 6 (2006) 82 To examne the relatve reacton of stock prces at the frst tradng day after rejecton of moton, three regresson models are run by usng dfferent hstorcal estmaton of betas as exploratory varables. Two of stocks are not ncluded, because ther systematc rsk coeffcents are not statstcally sgnfcant. The results of three separate regresson models are summarzed n Table 1. Table 1: Regresson Results of Monday Indvdual Stock Return on Hstorcal Betas for Three Dfferent Estmaton Perods (60 days, 120 days, 240 days) Model I (60) Model II (120) Model III (240) α -0. 0605-0.06134-0.0442 t statstc -3.5937*** -3.7904*** -2.6447** β -0.06478-0.0647-0.0849 t statstc -4.1724*** -4.2913*** -5.1884*** R 2 0.2789 0.2859 0.3655 Adjusted R 2 0.2588 0.2704 0.3555 F statstc 17.409*** 18.4154*** 29.92*** *,**,*** ndcate statstcal sgnfcance at 10%, 5% and 1%, respectvely. The results of all three regresson models suggest that hstorcally estmated systematc rsks of stocks have statstcally sgnfcant explanatory power for the percentage declne of stock prces on the day of sharp market fall, regardless of tme perods over whch hstorcal betas are estmated. Negatve slope coeffcents of regresson models are all statstcally sgnfcant and vary from -0.06 to -0.044. Negatve slope coeffcents as expected n case of market effcency ndcate that stocks wth hgher betas decrease more than stocks wth lower betas. Therefore, these fndngs confrm that stock market nvestors are able to reflect ratonally dfferent systematc rsk characterstcs of stocks on March 3, 2003. Although the results presented n Table 1 suggest that market behavor s ratonal n reflectng prevous nformaton about economc value of securtes, t can be sad that stock prces are overreact or underreact to new comng nformaton. If ths argument s correct, then predctable market adjustments are expected after the frst tradng day of event. To examne whether expected changes n value of stocks are less or more than true economc value, abnormal returns for each stock on the frst day of event were estmated by followng formula. AR = R ( α + β Rm ) (2) where: R = Daly returns to stock on the frst tradng day of event R m = Daly returns to the market ndex on the frst tradng day of event α = The market model ntercept estmated over perod t β = Systematc rsk of stock calculated over perod t t = 60 days, 120 days, and 240 days Then Stocks were ranked by ther abnormal returns on the frst tradng day after the event. Portfolo A ncludes frst 24 stocks havng lower abnormal returns and Portfolo B conssts of remander 24 stocks wth hgher abnormal returns on the frst day of events. Usng formula 2, the abnormal returns for each stock were then calculated for four subsequent days of week over whch market drecton was up.
83 Internatonal Research Journal of Fnance and Economcs - Issue I (2006) Table 2: Abnormal Returns of Portfolo A and Portfolo B after the Event PANEL A: Based on Market Model Estmated on 60-day Data Frst tradng day Subsequent Dates(market drecton for each day was up) after event 03.03.2003 04.03.2003 05.03.2003 06.03.2003 07.03.2003 Portfolo A -0.0108 0.0100 0.00170 0.0030 0.0069 Portfolo B 0.0363 0.0002 0.0009-0.0062-0.0022 Mean A-Mean B -0.0471 0.0098 0.0008 0.0092 0.0090 t stat. -10.157*** 1.317 0.323 1.920* 1.304 M-W stat. -5.938*** -1.876* -0.454-2.144** -1.217 PANELB: Based on Market Model Estmated on 120-day Data Frst tradng day Subsequent Dates (market drecton for each day was up) after event 03.03.2003 04.03.2003 05.03.2003 06.03.2003 07.03.2003 Portfolo A -0.0135 0.0091 0.0006 0.0011 0.0027 Portfolo B 0.0353 0.0028 0.0015-0.0041 0.0025 Mean A-Mean B -0.0488 0.0063-0.0009 0.0052 0.0002 t stat. -12.322*** 0.868-1.66 1.887 0.022 M-W stat. -5.938*** -1.505-0.124-1.794* -1.278 PANEL C: Based on Market Model Estmated on 240-day Data Frst tradng day Subsequent Dates (market drecton for each day was up) after event 03.03.2003 04.03.2003 05.03.2003 06.03.2003 07.03.2003 Portfolo A -0.0145 0.0105 0.0011 0.0016 0.0032 Portfolo B 0.0231 0.0061 0.0051-0.0044 0.0031 Mean A-Mean B -0.0376 0.0044-0.0040 0.0060 0.0001 t stat. -8.839*** 0.559-1.170 1.253 0.0260 M-W stat. -5.938*** -1.196-0.928-1.691* -1.361 *,**,*** ndcate statstcal sgnfcance at 10%, 5% and 1%, respectvely. To test whether there were any predctable patterns of abnormal returns for portfolos after the frst tradng days, the abnormal returns of portfolos are compared by usng t test. Besdes t-tests, Mann-Whtney test s also conducted as a nonparametrc test. Table 2 reports test results and means of portfolos. The dfference between abnormal returns of Portfolo A and that of Portfolo B constructed by stocks whch were ranked accordng to the magntude of abnormal returns on the frst tradng day s statstcally sgnfcant to test whether there were any market adjustments over the subsequent days of week. Evdence shown n Table 2 ndcates that generally there s no strong sgn for underreacton or overreacton of nvestors that volates EMH after the rejecton of government s moton. 5. Concluson Effcent Market Hypothess (EMH) s wdely tested proposton by researchers. An mplcaton of EMH s that market prces reflect all avalable nformaton and expectatons, and that any new nformaton s properly ncorporated nto prces wthout any delay. Sharp fall of securty prces durng the turbulent tmes s a very unque opportunty for testng valdty of EMH. Unexpected events can create more stress n the market, and market partcpants may lose ther ablty to asses ratonally the valuaton mplcatons of events. On March 1, 2003 Turksh parlament rejected the hghly controversal bll that allows the deployment of U.S troops n Turkey. The rejecton of the bll shook the poltcal and economc relatons between Turkey and U.S. Ths paper examnes the prcng behavor of the Turksh Stock Market after the surprsngly rejecton of the bll authorzng the deployment of U.S troops n Turkey. The emprcal results generally show that at the frst tradng day after rejecton of the moton, hstorcal estmaton of betas was hghly sgnfcant exploratory varables. Also, there s no clear sgn for underreacton or overreacton of nvestors that volates the assumptons of EMH.
Internatonal Research Journal of Fnance and Economcs - Issue 6 (2006) 84 References [1] Angboza, Lazarov A. and Ranga Narayanan (1996), Catastrophc Shocks n the Property- Lablty Insurance Industry: Evdence on Regulatory and Contagon Effects, The Journal of Rsk and Insurance, Vol. 63, No. 4, pp. 619-637. [2] Blson, Chrstopher M., Tmothy J. Bralsford and Vncent C. Hooper (2002), The Explanatory Power of Poltcal Rsk n Emergng Markets, Internatonal Revew of Fnancal Analyss, Vol. 11, Issue 1, pp. 1-27. [3] Blose, Laurence E., Robn Bornkamp, Marc Brer, Kends Brown, Jerry Frederck (1996), Catastrophc Events, Contagon, and Stock Market Effcency: The Case of the Space Shuttle Challenger, Revew of Fnancal Economcs, Vol. 5, No. 2, pp. 117-129. [4] Carter, Davd A. and Betty J. Smkns (2004), The Market s Reacton to Unexpected, Catastrophc Events: The Case of Arlne Stock Returns and the September 11th Attacks, The Quarterly Revew of Economcs and Fnance, Vol. 44, Issue 4, pp. 539-558. [5] Chen, Dar-Hsn, Feng-Shun Bn and Chun-Da Chen (2005), The Impacts of Poltcal Events on Foregn Insttutonal Investors and Stock Returns: Emergng Market Evdence From Tawan, Internatonal Journal of Busness, Vol.10, No. 2, pp. 165-188. [6] Damonte, Robn, John M. Lew, Ross L. Stevens (1996), Poltcal Rsk n Emergng and Developed Markets, Fnancal Analysts Journal, Vol. 52, Issue 3, pp. 71-76. [7] Fenberg, Martn and Damr Tokc (2002), Beta and Return: One-day Effect, Journal of Asset Management, July 2002, Vol. 3, No.1, pp. 67-72. [8] Glascock, John L., Davd Mchayluk, Karyn Neuhauser (2004), The Rskness of REITs Surroundng the October 1997 Stock Market Declne, Journal of Real Estate Fnance and Economcs, Vol. 28, Vol. 4, pp. 339-354. [9] Harrs, John M. and Mchael F. Spvey (1990), Systematc Prcng Durng the Stock Crash: Assessng Market Effcency, Journal of Busness Research, 21, pp. 59-68. [10] Hll, Joanne and Thomas Schneewes (1983), The Effect of Three Mle Island on Electrc Utlty Stock Prces: A Note, Journal of Fnance, Vol. 38, No. 4, pp. 1285-1292. [11] Kalra, Rajv, Glenn V. Henderson Jr. and Gary A. Ranes (1993), Effects of the Chernobyl Nuclear Accdent on Utlty Share Prces, Quarterly Journal of Busness and Economcs, Vol. 32, No. 2, pp. 52-77. [12] Kryzanowsk, Lawrence, Lorne Swtzer and L Jang (1995), Stock Market Crash Behavour of Screen-Sorted Portfolos, Internatonal Revew of Economcs and Fnance, Vol. 4, Issue 3, pp. 227-244. [13] Lamb, Renhold P. (1998), An Examnaton of Market Effcency Around Hurrcanes, The Fnancal Revew, Vol. 33, Issue 1, pp. 163-172. [14] Le Vu, Quan and Paul J. Zak (2006), Poltcal Rsk and Captal Flght, Journal of Internatonal Money and Fnance, Vol. 25, Issue 2, pp. 308-329. [15] Lmmack, R.J. and C. W. R. Ward (1990), The October 1987 Stock Market Crash: An Exploratory Analyss of Share Prce Models, Journal of Bankng and Fnance, Vol. 14, Issue 2-3, pp. 273-289. [16] Maloney, Mchael T. and J. Harold Mulhern (2003), The Complexty of Prce Dscovery n an Effcent Market: The Stock Market Reacton to the Challenger Crash, Journal of Corporate Fnance, Vol. 9, Issue 4, pp. 453-479. [17] Mehdan, Seyed, Tevfk Nas and Mark J. Perry (2005), An Examnaton of Investor Reacton to Unexpected Poltcal and Economc Events n Turkey, Frst Internatonal Conference on Bussness, Management and Economcs, 16-19 June, 2005, Cesme-İzmr, Turkey. [18] Pompan, Mchael M., John M. Longo (2004), A New Paradgm for Practcal Applcaton of Behavoural Fnance: Creatng Investment Programs Based on Personalty Type and Gender to Produce Better Investment Outcomes, The Journal of Wealth Management, Vol. 7, Issue 2, pp. 9-15.
85 Internatonal Research Journal of Fnance and Economcs - Issue I (2006) [19] Shelor, Roger M., Dwght C. Anderson and Mark L. Cross (1990), The Impact of the Calforna Earthquake on Real Estate Frms Stock Value, The Journal of Real Estate Research, Vol. 5, No. 3, pp. 335-340.