BEBR m FACULTY WORKING PAPER NO. 89-1595 The Intertemporal Relation Between U. S. and Japanese Stock Markets Kent G. Becker Joseph E. Finnerty Manoj Gupta College of Commerce and Business Administration Bureau of Economic and Business Research University of Illinois Urbana-Champaign BEBR FACULTY WORKING PAPER NO. 89-1595 College of Commerce and Business Administration University of Illinois at Urbana- Champaign September 1989 The Intertemporal Relation Between U. S and Japanese Stock Markets Kent G. Becker Temple University Joseph E. Finnerty University of Illinois Manoj Gupta University of Illinois Department of Finance ABSTRACT This paper finds a high correlation between the open to close returns in U. S. stocks in the previous trading day and the performance in the Japanese equity market in the current period. In contrast, the Japan- ese market has only a small impact on the U. S. return in the current period. High correlations among open to close returns are a violation of the efficient market hypothesis; however, in trading simulations, the excess profits in Japan vanish when transactions costs and transfer taxes are included. Digitized by the Internet Archive in 2011 with funding from University of Illinois Urbana-Champaign http://www.archive.org/details/intertemporalrel1595beck THE INTERTEMPORAL RELATION BETWEEN THE U.S. AND JAPANESE STOCK MARKETS The two largest stock markets in the world in terms of capitaliza- tion, volume, and shares listed are the Tokyo Stock Exchange (TSE) and the New York Stock Exchange (NYSE). Because Tokyo is 14 hours ahead of New York, there is an eight hour difference between the close of the TSE and open of the NYSE. Since there is no overlap between the two markets, traders or technical analysts may look to the TSE as a predictor of market movement on the NYSE and/or examine changes on the NYSE as indicators of performance on the TSE. , As shown in Figure 1, the TSE opens at 7:00 p.m. Eastern Standard Time (EST) and closes at 1:00 a.m. EST. The NYSE opens at 11:00 p.m. Japanese time (9:00 a.m., EST) and closes at 5:00 a.m. Japanese time (3:00 p.m., EST). Thus, there is no common time interval in which both markets are open. Insert Figure 1 about here High correlations between the respective open to close returns are a violation of the efficient market hypothesis because public infor- mation about the performance in one market could be used to profitably -2- trade in another market. If the markets are efficient, information about the open to close performance in one market (for example, the U.S. return in period t-1) will be fully reflected in the open price in the other market (Japan in period t, for example). Since new in- formation flows randomly into the market, subsequent price changes should be random and the open to close returns in Japan will be uncor- rected with the U.S. returns. Thus, the performance in the U.S. should affect the open price in Japan and the correlation between the open to close returns in the two markets will be zero. Early research on the synchronization among stock prices across countries [Grubel (1968), Levy and Sarnat (1970), Agraon (1972), Ripley (1973), Lessard (1976), Panton, Lessiq, and Joy (1976), and Hilliard (1979)] focused on the benefits of international diversification in 2 reducing systematic portfolio risk. Most of the studies used weekly or monthly return data for a number of years and found that correla- tions across countries were statistically insignificant or very low. Recent research on this topic investigated international equity market linkages. Using daily closing market prices for five coun- tries, Jaffe and Westerfield (1985a) found that correlations between the U.S. and the other market returns for each day of the week were generally positive and significant. Schollhammer and Sand (1985) and Eun and Shinr (1989) employed daily market closing data in the 1980s for several countries and discovered a substantial amount of inter- dependence among national stock markets. Bennett and Kelleher (1988), Dwyer and Hafer (1988), Goodhart (1988), King and Wadhwani (1988), Neumark, Tinsley, and Tosini (1988), -3- and Roll (1988) investigated international equity market linkages around the October 1987 crash. King and Wadhwani (1988) and Goodhart (1988) used hourly data for the New York, London, and Tokyo markets. They found strong cross-exchange linkages after the crash. Neumark, Tinsley, and Tosini (1988) focused on U.S. stocks that were also traded in London and Tokyo. Using opening and closing prices in New York and closing prices in London and Tokyo for eight months after the crash, they discovered that the predictive ability of after-hours pricing in foreign equity markets was strong after the crash, but de- clined sharply in later months. This paper employs opening and closing data for market averages in the U.S. and Japan for a longer time period, from 1985 to 1988, to study the synchronization of stock price movements. Therefore, our focus is not the transmission of stock prices and volatility during the crash. There are two advantages of opening and closing prices over only closing data. First, direct tests of market efficiency can be conducted in which a simulated trader in Japan buys or sells at the opening price, depending on the performance in the U.S. market the previous day. Second, the influence of the daily return in one market on the overnight return of the other market can be investigated. The results indicate that the performance in the U.S. greatly in- fluences open to close stock returns in Japan the next day and the change in the TSE has only a slight impact on the NYSE performance the same day. Large movements in the U.S. predict open to close returns in Japan the next day remarkably well. However, when Japanese trans- actions costs and taxes are included, the excess returns from following -4- the U.S. vanish. In addition, the overnight return in Japan is greatly affected by the U.S. performance. In contrast, the Japan open to close return does not have an impact on the U.S. overnight return. I. DATA AND METHODOLOGY Daily opening and closing data for the Nikkei Index, S&P 500, and the yen/dollar exchange rate from October 5, 1985 to December 28, 1988 were obtained. It is believed that this period is more meaningful than a longer time period because of the structural changes in both the U.S. and Tokyo markets. Data for the Nikkei Index, which is a price-weighted index of 225 stocks on the TSE , were acquired from Nihon Keizai Shimbun (Japan Economic Journal). Opening and closing spot rates for the yen were gathered from the International Monetary Market Yearbooks. Arithmetic returns for the Nikkei Index and S&P 500 are calculated on a local and common currency basis. Common currency returns are computed by converting the opening and closing S&P 500 levels to yen equivalents. October 1987 was a very unusual period in the recent history of the stock market. To ensure that the results are not being driven by the data from the crash, two data sets are used in this study: the first with the crash month, October 1987, and the second without. Correlations between the open to close returns are computed for (1) TSE o-c and S&P o-c, which tests whether the Japanese market leads the U.S., and (2) TSE o-c and S&P , o-c, a test of the U.S. equities leading the Japanese. To determine how the open to close result in one market relates to the close to open in the other, the following -5- correlations are calculated: (1) S&P , o-c and TSE c-o to determine how the performance in the U.S. market affects the TSE close to open 3 returns and (2) S&P c-o and TSE o-c. Regressions are estimated to determine the relation between the two markets. As a test of the Japan market leading the U.S., a re- gression is estimated with TSE o-c as the independent variable and S&P o-c as the dependent variable. As a test of the U.S. leading Japan, a regression is run with S&P , o-c as the independent variable and TSE o-c as the dependent variable. Thus, the correlations and regressions are calculated on the local and common currency returns, with and without October 1987. In addition, simulated trading strategies are implemented on the data set without October 1987. In the simulation, a trader buys in the Japanese market when the local S&P 500 increases by .5%, 1%, 1.5% or 2%, the previous day and sells when the index decreases by the same 4 percentages. The positions are closed at the end of the day. Re- turns with round-trip transactions costs of 0%, .50%, and 1% are cal- culated. Profitable trading days are counted along with mean returns. II. EMPIRICAL RESULTS The results show that the performance in the U.S. strongly in- fluences Japanese returns while the Japan market has only a slight impression on the S&P 500. Tables 1 and 2 present the regressions and correlations between the open to close returns. The correlation be- tween the Nikkei and S&P 500 return in the current period, which is a test of the Japanese market leading the U.S., is significant for the -6- local returns with and without October 1987, and the common currency returns for the whole data set. The correlations range from .0463 to .1171. Thus, the Japanese performance in the current period explains only about one percent of the fluctuations in the U.S. returns. Insert Tables 1 and 2 about here In contrast, the performance in the U.S. in the previous trading day has a major impact on the Japanese return in the current day. All correlations between the lagged U.S. return and the current Nikkei re- turn, which range from .2667 to .4963, are significant at a 1% level. Again, the correlations for the common currency returns are slightly lower than the local currency returns. In addition, the correlations for the entire data set are much higher than for the data set without the crash month. As expected, the open to close returns in the U.S. affect the close to open in Japan. From Tables 3 and 4, correlations between the lagged U.S. return open to close and the Nikkei close to open returns are all significant at a 1% level, ranging from .3407 to .4205. In addition, the Japanese open to close performance does not have an im- pact on the U.S. overnight return. This result is surprising, since the same Japan daily return has a slight impact on the subsequent daily U.S. return. In effect, information which affects the Japanese market has little or no influence on the U.S. market. Insert Tables 3 and 4 about here -7- For the data sets without the crash month, the lagged U.S. return has more impact on the overnight Japanese return than on the following open to close return. When October 1987 is included, the correlation between the lagged U.S. performance and the Tokyo daily return is higher. The simulated trading strategies, presented in Table 5, reveal that, in the absence of transactions costs, the filter rules do a remarkable job of predicting up and down returns in Japan. The up triggers predict profitable trading days 72% to 81% of the time. Looser up triggers are better able to predict profitable Japanese trading days, with the exception of the 2% method. Insert Table 5 about here The down triggers foretell negative returns the next day with slightly less precision, 59% to 75%. Similarly, the looser down triggers are more effective at detecting negative returns. Whatever method is implemented, the resulted demonstrate that next day market performance in Japan is predicted by the U.S. market performance. However, the presence of transactions costs and taxes eliminate the profits and predictive ability of the filters. Trading costs are higher in Japan than in the U.S., over 1% round-trip for large insti- tutions when commissions and taxes are included. Table 5 presents mean returns for the various triggers; following the U.S. is not profitable when transactions costs are 1%, with profitable trading days well below 50%. When transactions costs are .5%, the percentage of profitable trading days is about 50% for the up 1.5% and 2% triggers. -8- Although the lagged U.S. return predicts performance in Japan re- markably well, it is impossible to profit from following the U.S. be- cause of the high trading costs in Japan. III. CONCLUSION From October 5, 1985 to December 28, 1988, the performance of the U.S. market had a great impact on Japanese equities. The S&P 500 re- turns in the previous day explain from 7-25% of the fluctuations in the Nikkei Index open to close returns the next day, demonstrating that the U.S. market greatly influences Japan. In addition, the per- formance in the U.S. in the previous day explains between 11-18% of the fluctuations in the Japan overnight returns. In contrast, the Japanese market has a small impact on U.S. equities, explaining only one percent of the fluctuations in U.S. open to close returns. Although this result is statistically sig- nificant, it is probably not high enough to profitably trade on in the U.S. In addition, there is no relation between the performance of the Japanese market and the close to open return in the U.S. Trading simulations are performed on the Japanese market based on U.S. performance. Various filters are implemented; all are successful in selecting profitable Japanese trading days with great regularity. However, high trading costs in Japan prevent Japanese arbitrageurs from profiting from this strategy. -9- FOOTNOTES The TSE takes a lunch break from 11:00 a.m. to 1:00 p.m. Tokyo time. 2 See Madura (1985) for a review of literature dealing with the co-movement of international stock prices, particularly in an equity portfolio context. 3 When prices could not be obtained for a lagged or current trading day due to a closed market in one country, the observation is deleted from the sample. For example, assume that both markets are open Thursday, Friday, and Monday, and the Japanese market is open Saturday For the test of Japan leading the U.S., returns are taken from Thursday, Friday, and Monday. For the test of the U.S. leading Japan, observations are taken from Friday and Saturday. A Monday return could not be calculated because the U.S. market was not open Saturday. For the U.S. affecting Japanese overnight returns, observations are taken for Friday and Saturday. Only the Friday U.S. overnight return is obtained for the test of Japan on the U.S. close to open returns. 4 Inclusion of the crash month would have yielded more dramatic re- sults because the S&P 500 open to close return decreased by 20.43% on October 19, followed by a fall of 14.90% in Japan the next day. The U.S. return increased 5.34% on October 20 and the Nikkei Index fol- lowed by rebounding 9.29%. After the S&P 500 declined by 20.43% on October 19, 1987, the overnight return in Japan was -.0066%. This outlier affects the re- sults. If only this return is deleted, the correlation between the lagged common currency U.S. returns and the Japanese open to close is .3846 and .3945 for the local currency U.S. returns. For the time period of this study, 56.6% of the U.S. open to close returns were up (464 U.S. up trading days and 356 down). In Japan, 58.5% of open to close returns were up (522 returns greater than zero and 371 down). -10- The scale of commission rates established by the TSE is set out below: Value of Transaction (in Yen) less than 1,000,000 Yen 1,000,001 to 3,000,000 3,000,001 to 5,000,000 5,000,001 to 10,000,000 10,000,001 to 30,000,000 30,000,001 to 50,000,000 50,000,001 to 100,000,000 100,000,001 to 1,000,000,000 over 1,000,000,000 One Way % (of Commission Rate highest value) 1.2% 1.20% 1.00% + 2,000 Yen 1.07% .80% + 5,000 .90% .75% + 12,500 .88% .60% + 27,500 .69% .40% + 87,500 .58% .25% + 182,500 .41% .20% + 212,500 .22% .15% + 712,500 .17% for 3 Billion Yen In addition, a transactions tax between .18% to .50% is imposed on the seller. -11- REFERENCES Aderhold, Robert, Christine Cumming, and Alison Harwood, 1988, Inter- national linkages among equities markets and the October 1987 market break, Federal Reserve Bank of New York Quarterly Review , 34-46. Agmon, Tamir, 1972, The relations among equity markets: A study of share price co-movements in the United States, United Kingdom, Germany, and Japan, Journal of Finance 27, 839-855. Bennett, Paul and Jeanette Kelleher, 1988, The international transmis- sion of stock price disruption in October 1987, Federal Reserve Bank of New York Quarterly Review , 17-33. Dwyer, Gerald P. and R. W. Hafer, 1988, Are national stock markets linked? Federal Reserve Bank of St. Louis Review , 3-14. Eun, Cheol S. and Bruce G. Resnick, 1986, Estimating the correlation structure of international share prices, Journal of Finance 41, 313-330. Eun, Cheol S. and Bruce G. Resnick, 1988, Estimating the dependence structure of share prices: A comparative study of the United States and Japan, Financial Review 23, 313-330. Eun, Cheol S. and Sangdal Shim, 1989, International transmission of stock market movements, Journal of Financial and Quantitative Analysis 24, 241-256. Finnerty, Joseph E. and Thomas Schneeweis, 1979, The comovement of international asset returns, Journal of International Business Studies 10, 66-78. Goodhart, Charles A. E., 1988, The international transmission of asset price volatility, Financial Market Volatility (Federal Reserve Bank of Kansas City) . Grubel, Herbert G., 1968, Internationally diversified portfolios: Welfare gains and capital flows, American Economic Review 58, 1299-1314. Hilliard, Jimmy E., 1979, The relationship between equity indices on world exchanges, Journal of Finance 34, 103-114. Jaffe, Jeffrey and Randolph Westerfield, 1985a, The week-end effect in common stock returns: The international evidence, Journal of Finance 40, 433-454. -12- Jaffe, Jeffrey and Randolph Westerfield, 1985b, Patterns in Japanese common stock returns: Day of the week and turn of the year effects, Journal of Financial and Quantitative Analysis 20, 261-272. King, Mervyn A. and Sushil Wadhwani , 1988, Transmission of volatility between stock markets, London School of Economics Financial Markets Working Paper. Lessard, Donald R., 1976, International diversification, Financial Analysts Journal 32, 32-38. Levy, Haim and Marshall Sarnat, 1970, International diversification of investment portfolios, American Economic Review 60, 668-675. Madura, Jeff, 1985, International portfolio construction, Journal of Business Research 13, 87-95. Neumark, David, P. A. Tinsley, and Suzanne Tosini, 1988, After-hours stock prices and post-crash hangovers, Federal Reserve Board Working Paper. Panton, Don B., V. Parker Lessiq, and Maurice Joy, 1976, Co-movement of international equity markets: A taxonomic approach, Journal of Financial and Quantitative Analysis 11, 415-432. Ripley, D., 1973, Systematic elements in the linkage of national stock market indices, Review of Economics and Statistics 55, 536-361. Roll, Richard, 1988, The international crash of October 1987, Financial Analysts Journal 44, 19-35. Schollhammer , H. and 0. Sand, 1985, The interdependence among the stock markets of major European countries and the United States: An empirical investigation of interrelationships among national price movements, Management International Review 25, 17-26. D/363 MIDNIGHT Figure 1 Trading Hours for TSE and NYSE JAPAN DAY 1 6 a.m. NOON 6 p.m. ******** ******** TSE OPEN MIDNIGHT 10 a.m. ******** 4 p.m. U.S. DAY 1 10 p.m. 4 a.m, 10 a.m. ******** NYSE OPEN MIDNIGHT JAPAN DAY 2 6 a.m. NOON 6 p.m. ******** ******** TSE OPEN MIDNIGHT Table 1 Regression and Correlation Results-Local Currency Returns T Value in Parentheses S&P o-c = oc + BJSE o-c + £j Oct. 87 Included Oct. 87 Excluded INTERCEPT (a ) .0007 (1.34) .0010*** (2.58) TSE t (B ) .1475*** (3.27) .1000** (2.28) F VALUE R SQUARE C0RR 10.71 .0137 .1171*** 5.18 .0069 .0829** TSE t " c = a us + e us s&p t-i°" c + e us Oct. 87 Included Oct. 87 Excluded INTERCEPT (o us ) .0009** (2.51) .0010*** (3.29) S&P t (B us ) .3894*** (15.9) .2270*** (7.87) F VALUE R SQUARE CORR 253.60 .2453 .4963*** 61.89 .0759 .2756*** *** Significant at a 1% level ** Significant at a 5% level * Significant at a 10% level Table 2 Regression and Correlation Results-Common Currency Returns T Value in Parentheses S&P o-c = a, + B T TSE o-c + e „ t J J t J Oct. 87 Included Oct. 87 Excluded INTERCEPT (a ) .0003 (0.48) .0006 (1.39) TSE t < Bj > .1226** (2.49) .0642 (1.27) F VALUE R SQUARE C0RR 6.20 .0080 .0893** 1.61 .0021 .0463 TSE o-c = a TTO + 8 TTO S&P , o-c + e ~US w US"~~t-l Oct. 87 Included US Oct. 87 Excluded INTERCEPT (a ) U o .0010*** (2.86) .0011*** (3.51) s&p t (S US ) .3404*** (15.1) .1885*** (7.60) F VALUE R SQUARE CORR 229.49 .2280 .4775*** 57.83 .0711 .2667*** *** Significant at a 1% level ** Significant at a 5% level * Significant at a 10% level Table 3 Regression and Correlations Results-Local Currency Returns T Value in Parentheses S&P o-c = ct T + S T TSE o-c + e INTERCEPT (a ) TSE (0 ) I— «J F VALUE R SQUARE CORR ~ ~ „j K 'J " t J Oct. 87 Included Oct. 87 Exc -.0001 -.0002 (1.33) (1.33) .0097 .0001 (0.18) (0.05) .034 .000 .000 .000 .0071 .0002 TSE t o-c = a us + S us SSP t _ l0 -c + c^ Oct. 87 Included Oct. 87 Excluded INTERCEPT (a us ) .0001 (6.00) .0001 (5.62) s&p t (e us ) .0154*** (9.87) .0258*** (12.4) F VALUE R SQUARE CORR 97.55 .1161 .3407*** 154.41 .1768 .4205*** *** Significant at a 1% level ** Significant at a 5% level * Significant at a 10% level Table 4 Regression and Correlations Results-Common Currency Returns T Value in Parentheses S&P o-c - a, + 6 T TSE o-c + e T t J J t J Oct. 87 Included Oct. 87 Excluded INTERCEPT (a.) -.0003 (-1.35) -.0003 (-1.28) TSE t (8 ) .0071 (0.35) .0338 (1.24) F VALUE R SQUARE CORR .119 .0002 .0133 1.54 .0024 .0486 TSE o-c = a TTO + |3 Tt _S&P.. .o-c + e ~US " "US*""t-l Oct. 87 Included US Oct. 87 Excluded INTERCEPT (a ) U O .0001 (6.32) .0001 (6.17) s&p t (8 US ) .0142*** (9.94) .0213*** (11.7) F VALUE R SQUARE CORR 98.86 .1174 .3427*** 136.9 .1600 .4000*** *** Significant at a 1% level ** Significant at a 5% level * Significant at a 10% level Table 5 Performance of Nikkei Index in Day t Inclusive of Round Trip Transactions Costs (TC) After S&P 500 Local Return Trigger in Day t-1 (Data Without October 1987) 0% ROUND TRIP TC .50% 1% S&Pt-! UP by .5% (226 TIMES) MEAN TSE t RETURN % UP S&P t _i DOWN by .5% (151) MEAN TSE t RETURN % DOWN S&Pj--! UP by 1% (114) MEAN TSE t RETURN % UP S&Pt-i DOWN by 1% (83) MEAN TSE t RETURN % DOWN S&Pt-! UP by 1.5% (53) MEAN TSE t RETURN % UP S&Pt-! DOWN by 1.5% (46) MEAN TSE t RETURN % DOWN S&Pt-! UP by 2% (25) MEAN TSE t RETURN % UP S&Pt-! DOWN by 2% (20) MEAN TSE t RETURN % DOWN .341 72% -.160 37% -.658 17% .234 59% -.267 28% -.769 18% .507 78% .006 -.493 24% .350 63% -.150 35% -.652 25% .665 81% .163 60% -.337 30% .535 70% .036 43% -.465 30% .733 76% .230 60% -.270 36% .699 75% .201 40% -.299 30%