DOI 0.55/sbe-05-00 CO-MOVEMENTS OF U.S. AND EUROPEAN STOCK MARKETS BEFORE AND AFTER THE 008 GLOAL STOCK MARKET CRASH MERIC Ilhan Rder Unversty, Lawrencevlle, New Jersey, USA NYGREN Lan Ma Rder Unversty, Lawrencevlle, New Jersey, USA BENTLEY Jerome T Rder Unversty, Lawrencevlle, New Jersey, USA McCALL Charles W Rder Unversty, Lawrencevlle, New Jersey, USA Abstract: Emprcal studes show that correlaton between natonal stock markets ncreased and the benefts of global portfolo dversfcaton decreased sgnfcantly after the global stock market crash of 987. The 987 and 008 crashes are the two most mportant global stock market crashes snce the 99 Great depresson. Although the effects of the 987 crash on the comovements of natonal stock markets have been nvestgated extensvely, the effects of the 008 crash have not been studed suffcently. In ths paper we study ths ssue wth a research sample that ncludes the U.S stock market and twenty European stock markets. We fnd that correlaton between the twenty-one stock markets ncreased and the benefts of portfolo dversfcaton decreased sgnfcantly after the 008 stock market crash. Key words: European stock markets, portfolo dversfcaton, 008 stock market crash. Introducton Studyng stock market crashes has been one of the most popular research topcs n fnance. Wang et al. (009) study the determnants of stock returns n stock market crashes. Uygur et al. (05) nvestgate whch stocks lose more value n stock market crashes. The mpact of stock market crashes on the co-movement patterns of - 83 -
natonal stock markets has been studed extensvely [see, e.g., Arshanapall and Doukas (993); Merc and Merc (997)]. The portfolo dversfcaton mplcaton of the co-movements of natonal stock markets has long been a popular research topc n fnance. Low correlaton between natonal stock markets s often presented as evdence n support of the beneft of global portfolo dversfcaton [see, e.g., Levy and Sarnat (970); Solnk (974); Watson (978); Merc and Merc (989); DeFusco et al. (996)]. Events of global mportance tend to have sgnfcant mpact on the world's stock markets. Emprcal studes provde evdence that the co-movement patterns of natonal stock markets change sgnfcantly after stock market crashes. Arshanapall and Doukas (993), Lau and McInsh (993), and Merc and Merc (997) compare the pre-october-987 and post-october-987 perods and demonstrate that correlaton between natonal stock markets ncreased and global portfolo dversfcaton benefts to nvestors decreased sgnfcantly after the 987 crash. The 997-998 emergng markets fnancal crss s another global event that had a sgnfcant mpact on the world's stock markets. Merc et al. (000) and Yang et al. (003) provde emprcal evdence that the crss affected the co-movement patterns of the world's stock markets sgnfcantly and that the benefts of portfolo dversfcaton to global nvestors wth the emergng stock markets decreased consderably after the 997-998 crss. Hon et al. (004), Merc and Merc (004), and Merc et al. (007) study whether the September, 00 terrorst attacks n the Unted States affected the comovements of global stock markets by comparng the pre-september and post- September perods. They conclude that correlaton between natonal stock markets ncreased and the benefts of global portfolo dversfcaton decreased sgnfcantly after September, 00. The 987 and 008 crashes are the two most mportant global stock market crashes snce the 99 Great Depresson. Although the effects of the 987 stock market crash on the co-movements of natonal stock markets have been studed extensvely, the effects of the 008 stock market crash have not yet been studed suffcently. In a recent paper, Merc et al. (05) study the co-movements of the world s stock markets before and after the 008 stock market crash by usng the Prncpal Components Analyss methodology. They fnd that the movements of the world s major stock markets have become sgnfcantly closer and the beneft of global portfolo dversfcaton has decreased sgnfcantly after the 008 stock market crash. In ths paper, we make a contrbuton to the lterature on ths subject by comparng the co-movements of the U.S. and European stock markets before and after the 008 stock market crash. In addton to Prncpal Components Analyss, we also use a maxmum lkelhood method to test the equalty of the covarance matrces of natonal stock market returns for the pre- and post-crash perods to determne f there are sgnfcant changes n the co-movement patterns of the U.S. and European Stock markets after the 008 stock market crash. - 84 -
. The Effects of the 008 Stock Market Crash on the U.S. and European Economes The growng uncertanty n fnancal markets began to mpact the global economy n 008 as reflected by the decrease n the growth rate of real world output from 3.9 percent n the pre-crash 003-007 perod to.5 percent n 008. Table below shows real annual growth rates for the world economy and dfferent regons around the world from 003 to 03. The ongong uncertanty n fnancal markets along wth bank falures around the world, frozen credt markets, and sharply fallng real estate and other asset prces led to a percent declne n world output n 009, the frst worldwde contracton snce 946. The one-year decrease n global trade of 5 percent n 009 was the sngle largest declne n world trade snce the end of World War II (00 CIA World Factbook). Column four n Area Table : Annual Growth Rates n Real GDP Average Annual Real GDP Growth Rate (%) 003-007 Annual Real GDP Growth Rate (%) 008 009 Average Annual Real GDP Growth Rate (%) 00-03 World 3.9.5 -.5 South Amerca 6. 4.9-0.3 3.4 Latn Amerca 5. 3.5 -.6 3. Central Amerca 4..7-4.3 3.3 Europe 3. 0.8-4.5 0.8 Eastern Europe 6.8 4.6-5.6.4 Northern Europe 3. -0.4-4.8.6 Southern Europe.5-0. -4.6 -. Western Europe.5-4. Northern Amerca.9 - -.8. Asa 6 3.3. 4.3 Chna 9.6 9. 8. Afrca 6 5.4 3 3 Australa& New Zealand 3.4.3.7.9 Source: http://unstats.un.org/unsd/snaama/selbascfast.asp Table ndcates that the fnancal crss that began n the U.S. n 008 spread to the rest of the world by 009 as annual growth rates for all regons of the world fell that year. - 85 -
However, the mpact of the fnancal crss contagon around the world was uneven. About one-thrd of world output was produced by countres that contnued to expand n 009 and throughout the post-crash perod; whle about two-thrds of world output came countres that were n a recesson n 008. One factor that explans these regonal dfferences s the degree of captal market ntegraton. For example, emergng Latn Amercan economes wth strong growth prospects attract nternatonal nvestments and become more vulnerable to fnancal contagon durng a perod of crss. Dufrenotet et. al (0) found the 007 U.S. subprme fnancal crss was transmtted to Latn Amerca and contrbuted to stock market volatlty n the regon. Usng data from a wder tme perod, 006 to 00,Hwang (04) also found evdence of fnancal connectons between U.S. and Latn Amercan economes. Countres and areas of the world wth a low degree of captal market ntegraton would be less vulnerable to fnancal contagon. Ths could partally explan why growth n Afrca declned n 009, but the area dd not fall nto recesson. A second factor nfluencng dfferences n the regonal effects from the fnancal crss s related to countres that are n the process of transton from central plannng to market orented economes. As countres begn the transton process toward economc lberalsm and begn to be more ntegrated nto the global economy, they become more open to captal flows and become more dependent on foregn trade. Ths ncreased partcpaton n the global economy wth expandng export sectors and ncreased dependence on captal nflows can facltate economc growth but t can also make transton economes more susceptble to the negatve effects of contagon durng a fnancal crss. Stgltz (l998) argues that the contagon rsk facng transton economes can be mtgated to some extent by ensurng an approprate system of regulaton s establshed along wth the movement toward developng compettve markets. Shostya (04) examned the mpact of the 007-009 fnancal crss on the twenty-eght countres of the former Sovet Unon and found that the degree of transton and the sgnfcance of trade wth the EU ncreased the contagon experenced by these countres. In support of ths fndng, Table shows the 5.6 percent real GDP contracton n Eastern Europe n 009 was deeper than any other regon n the world. A thrd reason for the regonal dfferences resultng from the fnancal crss s due to the polcy responses that were mplemented around the world. Most countres and central banks around the world mplemented expansonary monetary and fscal polces to stmulate ther economes to offset the recessonary pressure resultng from the fnancal crss. The wdespread use of expansonary fscal polcy around the world caused most governments to run budget defcts ( 90 percent of countres experenced growng budget mbalances n 009 (CIA, 00)). By 0 concern about budget defcts had shfted the polcy consensus on how to deal wth the fnancal crss. About half the countres n the world contnued to use expansonary fscal and monetary polces, about 5 percent used restrctve fscal and monetary polces, and about 5 percent used a mx of expansonary and contractonary fscal and monetary polces (CIA, 03-04). The average growth rate n 0 for countres that contnued usng expansonary fscal and monetary polces was 4.9 percent; countres that shfted to - 86 -
restrctve fscal and monetary polces realzed an average growth rate of just 0.8 percent that year (CIA, 03-04). Gven the budgetary rules establshed by the EU, defcts were a sgnfcant concern n Europe. Balout packages that were provded to countres n Southern Europe were coupled wth strct austerty measures. Ths fscal austerty explans n part why the Southern Europe economy contnued to contract n the post-crash perod and why Europe had the lowest growth rate n the 00-03 perod compared to other regons n the world as shown n Table. Table shows pre- and post-crash economc data for the U.S. and the European countres ncluded n the statstcal analyses that follow. Annual GDP growth rates fell n all European countres durng the 009-03 fve-year post-crash perod. The average annual European post-crash growth rate, -0.3 percent, s sgnfcantly lower than the 3.4 percent average annual rate that occurred durng the 003-007 fve-year pre-crash perod. The t-test for equalty of mean annual European growth rates for the two perods s assocated wth a p-value<0.000000. Country Table : Selected Economc Data for the Europe and the U.S. GDP Growth Rates (%) a 003-007 009-03 Exports Imports (% GDP) b (% GDP) b 003-007 - 87-009- 03 003-007 009-03 Inflaton Rate (% change n CPI) a 003-009- 007 03 Austra.5 0.43 48.69 5.6 45.44 48.3.87.0 Belgum.37 0.36 73.7 78.38 68.97 77.05.0.9 Czech Rep. 5.47-0.45 59.73 70.07 58.30 65.74.04.8 Denmark.0-0.7 47.39 5.5 4.47 46.0.73.9 Fnland 3.58 -.6 40.66 38.38 35.55 38.38.0.78 France.99 0.33 6.44 6.86 6.60 8.7.83.3 Germany.6 0.44 38.04 43.9 3.90 37.96.6.40 Greece 4.35-5.74 0.88 5.0 3.06 3.65 3..95 Hungary 3.45-0.83 66.60 84.8 68.73 78.7 5.34 4.09 Ireland 4.7-0.85 78.74 98.8 66.9 79.5 3.38-0.6 Italy.3 -.5 5.3 6.9 5.9 6.49.6.85 Netherlands.47-0.59 66.00 75.64 58.30 66.88.56.95 Norway.49 0.73 43.4 40.45 8.45 8.06.50.74 Poland 5..90 36.00 4.47 38.6 43.00.99 3.07 Portugal.3 -.53 8.33 33.57 36.37 37.6.70.44 Russa 7.50.03 33.75 9.08.0.6.7 7.73 Span 3.60 -.5 5.7 7.77 9.77 7.3 3.5.7 Sweden 3.5 0.87 45.40 45.49 38.7 40.33.6 0.89 Swtzerland.8. 54.4 65.33 45.86 55.48 0.88-0.09 Turkey 6.90 3.74.68 4.09 6.3 9.50.7 7.53 U.K. 3.03 0.30 5.65 9.33 8. 3.7.88 3.06 U.S..87.5 0.6.8 5.5 6.0.88.59 Source: World Bank Databank. a Effectve annual rate over ndcated fve year perod. b Average computed over ndcated fve year perod.
The effect of the crash on varatons n growth rates among the European economes mght shed lght on how the crash affected opportuntes for dversfcaton. The post-crash varaton n annual European growth rates s larger than the pre-crash varaton, but the dfference s not statstcally sgnfcant. The rato of post-crash to precrash varances n annual European growth rates s.9; however, the F-test for equalty of varances s assocated wth a p-value = 0.35. We emphasze that ths result s suggestve but not conclusve evdence of opportuntes for dversfcaton. Varatons n annual growth rates do not reflect ntra-country varatons n stock prces or ntertemporal co-movements of stock prces. Austra Belgum Country Czech Republc Denmark Fnland France Germany Greece Hungary Ireland Italy Netherlands Norway Poland Table 3: Major Tradng Partners for Europe and the U.S. Export and Import Partners Export Partners: Germany 9.3%, Italy 6.3%, Swtzerland 5.% (03 est.) Import Partners: Germany 40.4%, Italy 6.%, Swtzerland 5.4% (03 est.) Export Partners: Germany 8%, France 6.%, Netherlands 3% (0) Import Partners: Netherlands 0.9%, Germany 4.%, France 0.6% (0) Export Partners: Germany 3.8%, Slovaka 9.%, Poland 6.% (0) Import Partners: Germany 9.5%, Poland 7.7%, Slovaka 7.4% (0) Export Partners: Germany 5.9%, Sweden 3.5%, UK 9.6% (0) Import Partners: Germany.%, Sweden 3.5%, Netherlands 7.5% (0) Export Partners: Sweden.%, Russa 9.9%, Germany 9.3% (0) Import Partners: Russa 7.7%, Sweden 4.8%, Germany 3.9% (0) Export Partners: Germany 6.7%, Belgum 7.5%, Italy 7.5% (0) Import Partners: Germany 9.5%, Belgum.3%, Italy 7.6% (0) Export Partners: France 9.%, US 7.9%, UK 6.5% (03 est.) Import Partners: Netherlands.9%, France 7.6%, Chna 6.3 (03 est.) Export Partners: Turkey.6%, Italy 9.9%, Germany 6.5% (03 est.) Import Partners: Russa 3.8%, Germany 9.5%, Italy 7.9% (03 est.) Export Partners: Germany 5.6%, Romana 6.%, Slovaka 6.% (0) Import Partners: Germany 5.%, Russa 8.8%, Chna 7.4% (0) Export Partners: US 7.9%, UK 7.3%, Belgum 5.6% (0) Import Partners: UK 39.8%, US 3.%, Germany 7.6% (0) Export Partners: Germany.6%, France.%, US 6.8% (03 est.) Import Partners: Germany 4.7%, France 8.4%, Chna 8.4% (03 est.) Export Partners: Germany 6.5%, Belgum 3.7%, France 8.8% (0) Import Partners: Germany 3.8%, Chna %, Belgum 8.4% (0) Export Partners: UK 5.6%, Germany.6%, Netherlands % (0) Import Partners: Sweden 3.6%, Germany.4%, Chna 9.3% (0) Export Partners: Germany 6%, UK 7%, Czech Republc 6.5% (0) Import Partners: Germany 7.3%, Russa.%, Netherlands 5.9% (0) Export Partners: Span.7%, Germany.4%, France.9% (0) - 88 -
Portugal Import Partners: Span 3%, Germany.5%, France 6.7% (0) Export Partners: Netherlands 4.6%, Chna 6.8%, Germany 6.8% Russa (0 est.) Import Partners: Chna 6.6%, Germany.%, Ukrane 5.7% (0 est.) Export Partners: France 6.8%, Germany 0.8%, Italy 7.7% (0) Span Import Partners: Germany.8%, France.5%, Italy 6.7% (0) Export Partners: Norway 0.4%, Germany 0.3%, UK 8.% (0) Sweden Import Partners: Germany 7.4%, Denmark 8.5%, Norway 8.4% (0) Export Partners: Germany 8.5%, US.6%, Italy 7.6% (03 est.) Swtzerland Import Partners: Germany 8.%, Italy 0.5%, France 8.5% (03 est.) Export Partners: Germany 8.6%, Iraq 7.%, Iran 6.5% (0) Turkey Import Partners: Russa.3%, Germany 9%, Chna 9% (0) Export Partners: Germany.3%, US 0.5%, Netherlands 8.8% Unted Kngdom (0) Import Partners: Germany.6%, Chna 8%, Netherlands 7.5% (0) Export Partners: Canada 8.9%, Mexco 4%, Chna 7.% (0) Unted States Import Partners: Chna 9%, Canada 4.%, Mexco % (0) Source: CIA World Factbook. Fgures are percentages of countres total exports or mports. Dates ndcate years of estmates. The deflatonary mpact of the crash s evdent as annual nflatons rates fell n 5 of the European countres durng the post-crash perod. Two countres, Ireland and Swtzerland, experenced deflaton over ths tme perod. However, the average annual pre-crash European nflaton rate of 3. percent s not statstcally dfferent from the average annual.3 percent post-crash rate. Also, the varance n annual European nflaton rates fell durng the post-crash perod, but not sgnfcantly. The data n Table show the heavy relance of the European economes on nternatonal trade, both before and after the crash. Exports n the European economes averaged 43. percent of ther GDPs durng the pre-crash perod, and 47.9 percent durng the post-crash perod. These fgures contrast sharply wth those for the U.S. economy. Durng the pre-crash perod, U.S. exports averaged just 0. percent and.8 percent of GDP, respectvely, n the pre- and post-crash perods. The mport data show comparable contrasts between the European and U.S. economes. The heavy relance of the European economes on trade s lkely to due to ther geographc proxmty and to the relatvely small szes of ther economes. Whle relance on trade may expose any sngle European country to rsk assocated wth adverse economc condtons suffered by ther tradng partners, t may present dversfcaton opportuntes for U.S. nvestors f European countres and the U.S. do not share major tradng partners. The data dsplayed n Table 3, whch lsts the top three tradng partners for the countres ncluded n our study, sheds lght on ths ssue. Canada, Chna, and Mexco are the three largest tradng partners of the U.S. Chna s a major tradng partner wth just 7 of the European countres, and no European country s a major tradng partner wth ether Canada or Mexco. - 89 -
Also, we note that just 5 of the European countres clam the U.S. as a major tradng partner. Much of the trade among the European countres appears to be nter- European. Germany, whch has the largest economy n Europe, s a major tradng partner wth all 0 other European countres n our sample. 3. Data and Methodology In addton to the U.S. stock market, the stock markets of the followng twenty European countres are ncluded n the study: Austra, Belgum, Czech Republc, Denmark, Fnland, France, Germany, Greece, Hungary, Ireland, Italy, the Netherlands, Norway, Poland, Portugal, Span, Sweden, Swtzerland, Turkey, and the U.K. We study and compare the co-movement patterns of the twenty-one stock markets before and after the 008 global stock market crash. We use the Morgan Stanley Captal Internatonal (MSCI) weekly U.S.-dollar stock market ndexes n the study. The data are drawn from the DataStream database. The weekly ndex returns are computed as the natural log dfference n the ndexes, ln (I I,t /I I,t- ). The pre-crash perod s the 003-007 fve year perod before 008 and the post-crash perod s the 009-03 fve year perod after 008. We frst compute and compare the correlaton coeffcents of the pre-crash and post-crash perods. We then use a maxmum lkelhood method to test the equalty of the pre- and post-crash covarance matrces to determne f there was a sgnfcant change n the covarance patterns of the twenty-one stock markets from the pre-crash to the post crash perod. We complete the study wth a prncpal components analyss of the co-movement patterns of the stock markets durng the pre- and post-crash perods. 4. Correlaton Analyss Table 4 shows the Pearson correlaton coeffcents between the U.S. stock market returns and the stock market returns of the twenty European stock markets durng the pre- and post-crash perods. All twenty correlaton coeffcents are hgher for the post-crash perod than for the pre-crash perod. The average correlaton coeffcent for all stock markets ncreased by 4 percent from 0.5 n the pre-crash perod to 0.737 n the post-crash perod. The correlaton between the U.S. and Hungaran stock markets ncreased by 83 percent from the pre-crash perod to the post-crash perod. These results ndcate that co-movements of the U.S. and European stock markets are substantally closer after the 008 stock market crash than they were before the crash. Ths mples that there are less portfolo dversfcaton opportuntes for U.S. nvestors wth European stock markets after the 008 stock market crash compared wth the precrash perod. The statstcs n Table 4 show that the U.S. stock market s most closely correlated wth the Dutch, French, German, and U.K. stock markets both before and after the 008 stock market crash (.e., these European stock markets are the worst - 90 -
portfolo dversfcaton prospects for U.S. nvestors n both perods). The U.S. stock market s least closely correlated wth the Portuguese, Polsh, Turksh, Czech, and Hungaran stock markets n both pre- and post-crash perods (.e., these European stock markets are the best portfolo dversfcaton prospects for U.S. nvestors n both perods). Table 4: Correlaton of the U.S Stock Market wth the Other Stock Markets Stock Markets Correlaton Coeffcents Pre-Crash Perod Post-Crash Perod % Change Netherlands 0.743 0.804 + 8 % France 0.73 0.87 + 3 % Germany 0.77 0.80 + 4 % U.K. 0.674 0.834 + 4 % Sweden 0.66 0.790 + 9 % Belgum 0.65 0.755 + 6 % Swtzerland 0.636 0.745 + 7 % Span 0.633 0.7 + % Italy 0.64 0.780 + 7 % Denmark 0.560 0.704 + 6 % Fnland 0.557 0.766 + 38 % Ireland 0.499 0.665 + 33 % Austra 0.460 0.78 + 70 % Norway 0.39 0.78 + 99 % Greece 0.39 0.80 + 09 % Portugal 0.380 0.647 + 70 % Poland 0.374 0.663 + 77 % Turkey 0.64 0.58 + 0 % Czech Republc 0.55 0.59 + 3 % Hungary 0.36 0.668 + 83 % Average 0.5 0.737 4 % Table 5 lsts the ten most correlated and ten least correlated pars of stock markets for the pre- and post-crash perods. Low correlaton coeffcents show the pars of stock markets wth the best portfolo dversfcaton beneft and hgh correlaton coeffcents show the pars of stock markets wth the least portfolo dversfcaton beneft. The Turksh stock market appears to be an attractve portfolo dversfcaton prospect for nvestors n both perods. Table 6 shows the average correlaton coeffcents of each stock market wth the other stock markets. A hgh average correlaton ndcates that a stock market s - 9 -
Studes n Busness and Economcs no. 0()/05 very well ntegrated wth the other stock markets. Such a stock market s not a good prospect for portfolo dversfcaton. Investors can maxmze the portfolo dversfcaton beneft by nvestng n the stock markets wth a low average correlaton coeffcent wth the other markets. The fgures n Table 5 ndcate that the Turksh, Czech, Hungaran, and Portuguese stock markets have the lowest average correlaton coeffcents n both pre- and post-crash perods (.e., these stock markets appear to be the best portfolo dversfcaton prospects for nvestors n both perods). The average of the average correlaton coeffcents ncreased by 6 percent from 0.587 n the pre-crash perod to 0.740 n the post-crash perod. Ths mples that global dversfcaton opportuntes wth the stock markets ncluded n ths study decreased substantally from the pre-crash perod to the post-crash perod. The average correlaton coeffcent ncreased the most for the Austran and Hungaran stock markets (54 percent) and the least for the Dansh stock market (8 percent). 5. Test of Equalty of the Covarance Matrces for the Pre- and Post-Crash Perods A modfed log-lkelhood rato test [see, e.g., Anderson (003)] s used to test the homogenety of the covarance matrces of the stock market returns of pre- and post-crash perods. Note that snce the sample correlaton matrx can be consdered as the covarance matrx under scale change, one can nterpret the test results as applyng to the correlaton matrces as well. Let Σ and Σ represent the covarance matrces of pre- and post-crash perods, respectvely. Each of the two random samples (one from each of the two dfferent hstorcal perods) conssts of 60 weekly Xequty market ndex returns of dfferent countres. Denote the sample data by, K, n, =, }, where p random X vector represents an observaton of the weekly stock market ndex returns of p = j dfferent countres from the th populaton and n = n = 60. Set n = n + n. The test statstc used to test for homogenety of Σ and Σ s p + 3p = = n n 6( p + ) n X X n = j j (for =,), A = ( n ρ log X X X λ, where ρ, () Σˆ = A = n (for =,), j= X j ˆ A + A Σ = n X )(, and X j X )' (for =,), () ( n ) Σˆ = λ = (3) ( n ) Σˆ - 9 -
Table 5: Most Correlated and Least Correlated Stock Markets Country Par Pre-Crash Perod Most Correlated Stock Markets Correlaton Coeffcent Post-Crash Perod Country Par Correlaton Coeffcent France Germany 0.943 France Germany 0.960 France Netherlands 0.96 France Netherlands 0.955 Germany Netherlands 0.90 France Italy 0.940 Belgum France 0.898 Germany Netherlands 0.936 France Italy 0.89 Italy Span 0.97 Belgum Netherlands 0.884 Italy Netherlands 0.93 France Swtzerland 0.883 France U.K. 0.90 France Span 0.880 Netherlands U.K. 0.90 France U.K. 0.874 France Span 0.900 Germany Span 0.87 Belgum Netherlands 0.900 Average 0.895 Average 0.93 Country Par Pre-Crash Perod Least Correlated Stock Markets Correlaton Coeffcent Post-Crash Perod Country Par Correlaton Coeffcent U.S. Hungary 0.36 Ireland Turkey 0.45 U.S. Czech Rep. 0.55 Greece Ireland 0.489 U.S. Turkey 0.64 Greece Turkey 0.500 Fnland Turkey 0.74 Ireland Czech Rep. 0.55 Fnland Hungary 0.34 Portugal Turkey 0.59 Swtzerland Turkey 0.35 Czech Rep. Greece 0.57 Ireland Turkey 0.36 Span Turkey 0.553 Fnland Czech Rep. 0.364 Greece Hungary 0.573 Netherlands Turkey 0.369 Czech Rep. Turkey 0.574 Germany Turkey 0.370 Swtzerland Turkey 0.576 Average 0.37 Average 0.58 Table 6: Average Correlaton of Each Stock Market wth the Other Stock Markets Stock Markets Average Correlaton Coeffcents Pre-Crash Perod Post-Crash Perod % Change (Least Integrated, Best Portfolo Dversfcaton Prospect) Turkey 0.47 0.587 + 4 % Czech Republc 0.459 0.65 + 4 % Hungary 0.46 0.7 + 54 % Portugal 0.509 0.77 + 4 % Fnland 0.509 0.753 + 48 % Poland 0.54 0.7 + 40 % Austra 0.55 0.795 + 54 % U.S. 0.5 0.7 + 38 % Greece 0.538 0.60 + % Ireland 0.56 0.656 + 7 % - 93 -
Norway 0.579 0.760 + 3 % Denmark 0.660 0.76 + 8 % Swtzerland 0.666 0.764 + 5 % Netherlands 0.668 0.85 + 4 % Span 0.669 0.758 + 3 % Belgum 0.670 0.770 + 5 % Italy 0.67 0.80 + 0 % U.K. 0.67 0.803 + 0 % Sweden 0.675 0.766 + 3 % Germany 0.68 0.80 + 0 % France 0.7 0.836 + 7 % (Most Integrated, Least Portfolo Dversfcaton Opportunty) Average 0.587 0.740 + 6 % Under the null hypothess of equal covarance matrces and multvarate normalty, the test statstc ρ log λ has approxmately a ch-squared dstrbuton wth degrees of freedom f = p( p + ) = 3. After mplementng the test, we fnd the value of the test statstc -ρlog λ 04.39 (roundng to three decmal places) wth p-value < 0-5. Snce the p-value s extremely close to zero, the test ndcates a sgnfcant dfference n the covarance matrces of Σ and Σ. As noted earler, the test results apply to the correlaton matrces as well. Hence we conclude that sgnfcant changes occurred n the correlaton patterns of the markets after the 008 crash. 6. Prncpal Components Analyss (PCA) We use the Prncpal Components Analyss (PCA) methodology to compare the co-movement patterns of the twenty-one stock markets durng the pre- and postcrash perods. Usng Keser s rule, statstcally sgnfcant prncpal components wth an egen value greater than unty are extracted for analyss. We use the Varmax rotaton to maxmze the factor loadngs of the stock markets n each prncpal component wth smlar movement patterns. The PCA technque has been used n several prevous studes to study the co-movements of natonal stock markets [see, e.g., Phlppatos et al. (983) Merc and Merc (989)]. A detaled dscusson of the PCA technque can be found n Marda et al. (979) and Marasculo and Levn (983). The PCA technque groups the stock markets n terms of the smlartes of ther movement patterns. The stock markets that are hghly correlated would have hgh factor loadngs n the same prncpal component. Therefore, nvestng n these stock markets can provde lmted portfolo dversfcaton beneft. The stock markets wth hgh factor loadngs n dfferent prncpal components have low correlaton. Therefore, an nvestor can maxmze portfolo dversfcaton beneft by nvestng n stock markets wth the hghest factor loadngs n dfferent prncpal components. Every stock market would have some factor loadng n each prncpal - 94 -
component. Some stock markets mght have hgh factor loadngs n more than one prncpal component. It ndcates that these stock markets are hghly correlated wth the stock markets wth hgh factor loadngs n more than one prncpal component. Such stock markets are not good prospects for global portfolo dversfcaton. Pre-Crash Perod Table 7 shows the factor loadngs of the prncpal components for the fve-year 003-007 pre-crash perod. There are two statstcally sgnfcant prncpal components n ths perod. The factor loadngs of the stock markets wth the hghest factor loadngs n each prncpal component are shown n bold. Some stock markets have hgh factor loadngs n the other prncpal component as well. These factor loadngs are shown n talcs and n lght font. Table 7: Prncpal Components Analyss: Pre-Crash Perod Stock Markets Prncpal Component # Prncpal Component # France 0.9 Netherlands 0.904 Germany 0.903 Belgum 0.855 Swtzerland 0.855 U.K. 0.89 Italy 0.87 Span 0.86 Sweden 0.805 U.S. 0.789 Denmark 0.673 0.548 Fnland 0.68 Ireland 0.68 Portugal 0.487 0.48 Hungary 0.835 Czech Republc 0.790 Poland 0.763 Turkey 0.699 Norway 0.44 0.696 Austra 0.503 0.688 Greece 0.505 0.50 Varance Explaned 45.8 % 5.7 % Cum. Var. Explaned 45.8 % 7.5 % The frst prncpal component explans 45.8 percent of the total varaton n the orgnal data matrx. The second prncpal component explans 5.7 percent of the total varaton n the orgnal data matrx. The two prncpal components together explan 7.5 percent of the total varaton n the orgnal data matrx. The frst prncpal - 95 -
component s domnated by the stock markets of relatvely well developed European countres and the U.S. stock market. The second prncpal component s domnated manly by the stock markets of relatvely less developed European countres. Investors can maxmze portfolo dversfcaton beneft by nvestng n the stock markets of the countres wth hgh factor loadngs n each prncpal component. The Dansh and Portuguese stock markets whch have hgh factor loadngs n the frst prncpal component also have hgh factor loadngs n the second prncpal component. The Norwegan, Austran, and Greek stock markets whch have hgh factor loadngs n the second prncpal component also have hgh factor loadngs n the frst prncpal component. Therefore, these stock markets are not good prospects for portfolo dversfcaton. Post-Crash perod Table 8 shows the factor loadngs of the stock markets for the post-crash perod. There s only one statstcally sgnfcant prncpal component n ths perod. Ths ndcates that all stock markets are hghly correlated and they are clustered n the same prncpal component n ths perod (.e., there are no stock markets wth sgnfcantly dfferent movement patterns from the other stock markets to justfy the creaton of a separate statstcally sgnfcant prncpal component). The prncpal component explans 75.7 percent of the total varaton n the orgnal data matrx. Table 8: Prncpal Components Analyss: Post-Crash Perod Stock Markets Prncpal Component # France 0.976 Netherlands 0.963 Germany 0.957 Italy 0.938 U.K. 0.93 Austra 0.99 Belgum 0.90 Sweden 0.898 Swtzerland 0.896 Norway 0.89 Span 0.889 Fnland 0.884 U.S. 0.849 Poland 0.845 Portugal 0.843 Denmark 0.84 Hungary 0.835 Ireland 0.770 Czech Republc 0.768 Greece 0.70 Turkey 0.693 Varance Explaned 75.7 % - 96 -
The movements of the French, Dutch, German, Italan, and U.K. stock markets are hghly correlated and they have the hghest factor loadngs n the prncpal component. The Irsh, Czech, Greek, and Turksh stock markets are relatvely less correlated wth the other stock markets. However, dversfyng nto these stock markets can provde only a non-sgnfcant portfolo dversfcaton beneft to nvestors. 7. Summary and Conclusons Emprcal studes demonstrate that correlaton between the world s stock markets ncrease and the benefts of global portfolo dversfcaton decrease after stock market crashes. 987 and 008 crashes are the most mportant global stock market crashes snce the 99 Great Depresson. Although the effects of the 987 crash on the co-movements of the world s stock markets have been studed extensvely, the effects of the 008 stock market crash have not been studed suffcently. We study ths ssue n ths paper by comparng the co-movement patterns of the U.S. and twenty European stock markets n the 003-007 fve-year pre-crash perod and the 009-03 fve-year post-crash perod. Our fndngs ndcate that the co-movements of the twenty-one stock markets changed sgnfcantly after the 008 stock market crash. The covarance matrx of the markets s sgnfcantly dfferent n the post-crash perod compared wth the pre-crash perod; the stock markets are more hghly correlated after the 008 crash than before the crash. Our prncpal components analyss results show that there s only one statstcally sgnfcant prncpal component n the post-crash perod compared wth two statstcally sgnfcant prncpal components n the pre-crash perod. Ths mples a closer co-movement pattern between the markets after the crash. These results ndcate that portfolo dversfcaton opportuntes n the U.S and European stock markets decreased sgnfcantly after the 008 crash. 8. References Anderson, T.W., (003), An Introducton to Multvarate Statstcal Analyss. Thrd edton. New Jersey: Wley & Sons. Arshanapall, B., Doukas, J., (993), Internatonal stock market lnkages: Evdence from the preand post-october 987 perod, Journal of Bankng and Fnance, Vol. 7, no., pp. 93-08. CIA World Factbook, 00, 03-04 DeFusco, R.A., Geppert, J.M., Tsetsekos, G.P., (996), Long-run dversfcaton potental n emergng stock markets, Fnancal Revew, Vol. 3, no., pp. 343-363. Dufrenot, G., Mgnon, V., Pegun-Fessolle, A., (0), The effects of the subprme crss on the Latn fnancal markets: An emprcal assessment, Economc Modelng, Vol. 8, no. 5, pp. 34-357. Hon, M.T., Strauss, J., Yong, S.K., (004), Contagon n fnancal markets after September : Myth or realty?, Journal of Fnancal Research, Vol. 7, no., pp. 95-4. - 97 -
Hwang, J.K., (04), Spllover effects of the 008 fnancal crss n Latn Amerca stock markets, Internatonal Advances n Economc Research, Vol. 0, no. 3, pp. 3-34. Levy, H., Sarnat, M., (970), Internatonal dversfcaton of nvestment portfolos, Amercan Economc Revew, Vol. 60, no. 3, pp. 668-675. Marasculo, L.A., Levn, J.R., (983), Multvarate Statstcs n Socal Scences: A Researcher's Gude, Monterey, Calforna: Brooks/Cole Publshng Company. Marda, K., Kent, J., Bbby, J., (979), Multvarate Analyss, New York, New York: Academy Press. Lau, S.T, McInsh, T.H., (993), Co-movements of nternatonal equty returns: A comparson of the pre- and post-october 9, 987, perods, Global Fnance Journal, Vol. 4, no., pp. -9. Merc, I., Echhorn, B.H., Nygren, L.M., Merc, G., (007), Co-movements of U.S., Japanese, and European equty markets before and after September, 00: Global portfolo dversfcaton mplcatons, Journal of Management Scence (Sprng), pp. -0. Merc, I., Merc, G., (989), Potental gans from nternatonal portfolo dversfcaton and ntertemporal stablty and seasonalty n nternatonal stock market relatonshps, Journal of Bankng and Fnance, Vol. 3, no. 4/5, pp. 67-640. Merc, I., Merc, G., (997), Co-movements of European equty markets before and after the 987 crash, Multnatonal Fnance Journal, Vol., no., pp. 37-5. Merc, I., Merc, G., (004), The dmnshng beneft of global portfolo dversfcaton, Global Busness and Fnance Revew, Vol. 9, no., pp. 43-5. Merc, I., Merc, G., Ratner, M., (000), Emergng market crss and the ASEAN-5 stock markets: 997-998, Internatonal Journal of Fnance, Vol., no. 3, pp. 835-853. Merc, I., Taga, L.S., Gshlck, H.E., Merc, G., (05), Co-movements of the world s stock markets before and after the 008 stock market crash, Journal of Strategc and Internatonal Studes, Vol. 0, no., pp. 74-8. Phlppatos, G.C., Chrstof, A., Chrstof, P., (983), The nter-temporal stablty of nternatonal stock market relatonshps: Another vew, Fnancal Management, Vol., no. 4, pp. 63-69. R Core Team (03). R: A Language and Envronment for Statstcal Computng. R Foundaton for Statstcal Computng, Venna, Austra. Avalable onlne at: http://www.r-project.org/ Shostya, A., (04), The effect of the global fnancal crss on transton economes, Atlantc Economc Journal Vol. 4, no. 3, pp. 37-33. Solnk, B.H., (974), Why not dversfy nternatonally?, Fnancal Analysts Journal, Vol. 30, no. 4, pp. 48-54. Stgltz, J., (988), More Instruments and Broader Goals: Movng Toward the Post Washngton Consensus, WIDER Annual Lecture, Helsnk. January, 05, avalable onlne at: http://www.wder.unu.edu/ publcatons/ annual-lectures /en_gb/al Uygur, O., Merc, G., Merc, I., (05), Whch stocks lose more value n stock market crashes?, Journal of Fnancal Plannng, Vol. 8, no., pp. 44-5. Wang, J., Merc, G., Lu, Z., Merc, I., (009), Stock market crashes, frm characterstcs, and stock Returns, Journal of Bankng and Fnance, Vol. 33, no.3, pp. 563-574. Watson, J., (978), A study of possble gans from nternatonal nvestment, Journal of Busness Fnance and Accountng, Vol. 5, no., pp. 95-06. Yang, J., Kolar, J.W., Mn, I., (003), Stock market ntegraton and fnancal crss: The case of Asa, Appled Fnancal Economcs, Vol. 3, no. 7, pp. 477-487. - 98 -