台灣股票市場的產業外溢效果 - 政大學術集成

全文

(1)國立政治大學國際經營與貿易學系碩士論文. 治政大 Stock Market Spillover of Industry Effect in Taiwan 立 ‧ 國. 學. 台灣股票市場的產業外溢效果. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. i n U. v. 指導教授：郭維裕博士 Advisor: Wei-Yu Kuo, Ph.D. 研究生: 張孟溢 Meng-Yi Chang 中華民國九十九年七月 July, 2010.

(2) Abstract We investigate the spillover of industry effect in Taiwan stock market. Using a generalized vector autoregressive where forecast-error variance decompositions are invariant to variable ordering, we objectively propose measures of both total and directional spillovers on return and volatility daily data. In full-sample analysis, there is a heavy spillover effect in the interaction between stock market and industries. The stock market acts as a receiver from the information diffused from the industries, but the industries could not be confirmed as spillover outputer or inputer. The. 政治大 Finally, conducting the robustness 立 test, we divide the sample periods into subperiods. rolling-sample findings also pinpoint the high spillovers during the financial events.. ‧ 國. 學. and switch the daily data toward weekly and monthly data, then obtaining the consistent results with prior inference.. ‧. n. al. er. io. sit. y. Nat. Keyword: Spillover, generalized VAR, industry, volatility, return. Ch. engchi. i. i n U. v.

(3) Contents Chapter 1. Introduction .................................................................................................. 1 Chapter 2. Generalized Spillover Definition and Measurement .................................... 5 Chapter 3. Data .............................................................................................................. 9 Chapter 4. Empirical study in Taiwan market.............................................................. 11 4.1 Full- sample spillover analysis ...................................................................... 12. 政治大 Chapter 5. Robustness .................................................................................................. 19 立 4.2 Rolling-sample spillover analysis .................................................................. 15. ‧ 國. 學. 5.1 Subperiods...................................................................................................... 19. ‧. 5.2 Weekly and monthly data ............................................................................... 22. sit. y. Nat. Chapter 6. Conclusion .................................................................................................. 25. n. al. er. io. References .................................................................................................................... 26. Ch. engchi. ii. i n U. v.

(4) 1. Introduction. Global capital market integration provides investors many ways to develop their diversified portfolio strategies and more and more empirical studies work on the interconnection across the international stock markets. King and Wadhwani (1990) and King, Sentana, & Wadhwani (1994) found that the international participants ignore fundamental economic information and simply focus on price movements in other countries. This may explain why it is difficult to explain the international. 政治大. correlation structure with macroeconomic variables while obvious comovements. 立. among world equity markets are observed. The literature focusing on the comovement. ‧ 國. 學. of international stock markets grew rapidly including Eun and Shim(1989), Hamao et al. (1990), Karolyi (1995), Richards (1995), Harvey (1995), Forbes and Rigobon. ‧. (2002), amd Morana and Beltratti (2006). Forbes and Rigobon (2002) show that. y. Nat. io. sit. correlation coefficients between countries are conditional on market volatility. After. n. al. er. the heteroskedasticity adjustment for the bias, there is a high level of market. i n U. v. comovement in all periods, which they call it interdependence, and that no contagion effect happened.. Ch. engchi. Recently, there are some literature discussing factors that drive the comovement between stock returns and volatility. Lesard (1974) suggested that country factors are the dominant drivers in security-price movement. Heston and Rouwenhorst (1994) found that industrial structure explains very little of the cross-sectional difference in country return volatility and the low correlation between country indices is almost completely due to country-specific sources of return variation. Conversely, Cavaglia, Brightman and Aked (2000) present evidence that industry factors have been grown in relative importance and may now dominate country factors. 1.

(5) Although the debate on the degree of importance of industry or country factors suspends, Hong, Torous and Valkanov (2007) propose an idea whether the returns of industry portfolios predict stock movements. 1Their investigation builds on Hong and Stein`s (1999) hypothesis of gradual diffusion of information across asset markets that leads to cross-asset return. They found that 14 industry portfolios in US are statistically significant in predicting the market by one or two months. These findings also suggest that stock markets react with a delay to information contained in industry returns about their fundamentals.. 政治大 the spillover of the stock returns and volatility across equity markets. The GARCH 立 In general, the concept of comovement covers not only the correlations but also. model put forth by Engle (1982) and Bollerslev (1986) was applied to specify the. ‧ 國. 學. heteroskedasticity of the intra-daily exchange rate, finding the evidence to support the. ‧. alternative hypothesis of intra-daily volatility spillovers from one market to the next.. sit. y. Nat. Hamao et al. (1990) explored the relationship between three markets New York,. io. er. London and Tokyo from an ARCH model and found significant spillovers of price and volatility. In this respect, a further category of model, Markov Switching Models,. al. n. v i n C h and Susmel 1994, is developed (Hamilton 1989, Hamilton Edwards and Susmel 2001, engchi U Edwards and Susmel 2003 ) to characterize the structure of the spillover transmission across markets. However, Diebold and Yilmaz (2009) provide another measure of interdependence of asset returns and volatility spillover. They focus on variance decompositions derived from vector autoregressive (VAR) model. The innovative construction of the. total spillover index could be continuously varying unlike the state indication of Edwards and Susmel (2001), while distracted from the definition and contagious. 1. Traditionally, many economists have recognized that investors rather than possessing unlimited processing capacity, are characterized as boundedly rational (Shiller, 2000; Sims, 2001). 2.

(6) issues such as Forbes and Rigobon (2002). Nonetheless, in order to solve the ordering problem of variables and put more emphasis on assets issue, Diebold and Yilmaz (2010) employed the generalized VAR model (Koop, Pesaran and Potter ,1996 ; Pesaran & Shin ,1998) which is invariant to the ordering toward variance decomposition on cross-asset spillover effect. In this paper, we choose the distinct VAR model from Diebold and Yilmaz (2010) to investigate the spillover effect among the main industries in Taiwan stock market. In the literature of the comovement in Taiwan, we provide a different point of view to. 政治大 widely discussed in the literature before. We would categorize the stocks into two 立 look into the relationship between industries rather than the country groups which was. main industries electronics and finance. These industries contain particular. ‧ 國. 學. information about economic activities ahead of the stock market (Hong, Torous and. ‧. Valkanov ,2007), so we attempt to capture the information diffusion effect by means. sit. y. Nat. of the variance decomposition.. io. er. The application on the variance decomposition of the generalized VAR model provides several advantages in our study. First, it is not a static model to depict the. al. n. v i n correlation within markets but aCdynamic model to analyze h e n g c h i U the shock into others with. the lagging-period order in variables. This nature could let us closely track the spillover effect in different periods. Second, this extended VAR model eliminates the problem of ordering of variables. Otherwise, the sequence of variables has a large influence on the variance decomposition results in traditional VAR models. Third, the variance decomposition distills the information into a single spillover measure. It provides an intuitive way to understand and calculate. Our findings show that there is a strong interaction between industries and stock market in the view of high spillover among them which is partly consistent with Hong, Torous and Valkanov (2007). Further, the stock market plays a role as a receiver to 3.

(7) take the spillover effect from industries, however we could not determine whether the industries to be a spillover inputer or outputer. In the middle of context, we also point out the critical financial event in the historical change of Taiwan stock market and industries spillover effect. The paper proceeds as follows. In section 2, we discuss the generalized VAR model and introduce the total and directional spillover indices. We describe our data in section 3 and present the empirical results in Taiwan stock market section 4. In section 5, we conduct the robustness analysis of our results. Finally, conclude in section 6.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 4. i n U. v.

(8) 2. Generalized Spillover Definition and Measurement. We measure the return and volatility spillovers based on the vector autoregressive (VAR) model developed by Sims (1980). However, the magnitude of the VAR model we conduct mainly focuses on the variance decompositions. We progress by measuring the total and directional spillovers in a generalized VAR model. Consider a covariance stationary N-variable VAR(p) p. x t    i x i1   t. (1). 政治大. i 1. where  ~ (0, ) . The moving average representation is. 立. . x t   Ai t i.   2 Ai 2  ...   p Ai  p. A 0 i. for i. 0 . The moving average. io. er. an N  N identity matrix and. 0. (3). y. i 1. Nat. A. 1. obey the recursion. sit. ‧ 國 i. i. ‧. A  A. A. 學. i 0. where the N  N coefficient matrices. with. (2). coefficients are the access to parse the forecast error variances of each variable into. al. n. v i n parts attributable to the variousC system shocks. Variance h e n g c h i U decompositions allow us to assess the fraction of the H-step-ahead error variance in forecasting shocks to. x. j. x. i. that is due to. , j  i , for each i.. The calculation of variance decompositions requires orthogonal innovations, but our VAR innovations are generally correlated. Although the Cholesky factorization achieve orthogonality, the variance decompositions then depend on ordering of the variables. The orderings could cause different statistical results. Under Koop, Pesaran and Potter`s (1996) and Pesaran & Shin`s (1998) framework on the generalized VAR which produces variance decompositions invariant to ordering, we could sidestep the 5.

(9) problem to reorder the variables. Hereafter, we call the generalized VAR model “KPPS Model”. First, we decompose the variance into both two parts: own variance share and cross variance shares. The own variance share is defined as the fraction of the H-step-ahead error variances in forecasting. x. i  1, 2,..., N . The cross. x , for. due to shocks to. i. i. variance shares, or spillovers, are defined as the fractions of the. x. H-step-ahead error variances in forecasting i, j  1, 2,..., N , such that j  i .. due to shocks to. i. x. j. , for. 政治大. Assume ijg ( H ) represents the KPPS H-step-ahead forecast error variance. 立.  ijg ( H ) . 1 ii. H 1. 學. H 1.  (e À  e ) h. i. h 0. 2. j.  (e À  A è ) h. i. h 0. h. ‧. ‧ 國. decompositions , for H = 1,2,…, then we have. i. n. Ch. j 1. N. i , j 1. g ij. ( H )  1 . After the reconstruction. i n U. ijg ( H ). v. (5). N.  j 1. . g ij. engchi. ijg ( H ) . g so that ij ( H )  1 and. . er. io. al. of the model, we adjust it as:. j 1. sit. y. Nat N. In general, it does not have to sum to one:. N. (4). g ij. (H ). (H )  N .. Last, we could use the volatility contributions from the KPPS variance decomposition to construct the volatility spillover index:. 6.

(10) N. N.  ijg ( H ). Total Spillover Index = S g ( H ) . i , j 1 i j N. . i , j 1. .  100  g ij. i , j 1 i j. g ij. (H )  100. N. (H ). (6). Moreover, we measure directional volatility spillovers received by market i from all other markets j as: N.  Directional Spillover Index (Received) = Sig ( H ) . j 1 i j N. 政治大 j 1. 立. g ij. (H )  100. g ij. (7). (H ). ‧ 國. 學. In a similar way we measure direction volatility spillovers transmitted by market i to all other market j as:. ‧. N. io. Ch. engchi. i n U. (H ). y. j 1. n. al. . g ji. er. Directional Spillover Index (Transmitted) = Sgi ( H ) . j 1 i j N. g ji. sit. Nat. .  100. (8). (H ). v. The above set of directional spillovers could be recognized as providing a decomposition of total spillovers into those coming from (or going to) a particular source. Finally, we acquire the net volatility spillovers transmitted form market i to all other markets j as Net Spillover Index = Sig ( H )  Sgi ( H )  Sig ( H ). (9). It is simply the difference between gross volatility spillover to and gross volatility received from all other markets. In the empirical work that follows, we use first-order. 7.

(11) generalized VAR with 10-step-ahead forecasts.2. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. 2. Ch. engchi. i n U. v. Our method is mainly derived from Diebold and Yilmaz (2010)`s generalized model which is consistent with our empirical results. 8.

(12) 3. Data. We use the generalized framework to measure the spillovers in Taiwan stock market and industry portfolios. The underlining daily data are obtained from the TEJ database from November, 1995 through April, 2010, for a total 3743 observations. The stock market variables we use are Taiwan weighted stock index (TAIEX) and Over The Counter Index, because these two capitalization weighted indices could truly reflect the capital market in Taiwan`s whole industries to test our hypothesis. On. 政治大 sectors to represent the industry 立 portfolios. The electronic sector constitutes about a. the ground of Taiwan`s unique industrial structure, we choose electronic and financial. ‧ 國. 學. fifty to seventy percent weight in TAIEX and over the courter index respectively, whereas the financial industry which is the second largest sector in Taiwan. So the. ‧. two leading industries could be utilized to consider the spillover effect in industry. sit. y. Nat. portfolios toward the stock markets. The Electronic Index and Finance Index are. n. al. er. io. compiled by Taiwan Stock Exchange (TSE). We use these four variables TAIEX,. i n U. v. Over the Counter Index, Electronic Index and Financial Index to construct our empirical model.. Ch. engchi. In general, we first classify the spillovers into two dimensions: return spillovers and volatility spillovers. To do so we could aggregate the spillover effect from two types of consecutive time series to get more detailed information among the stock market and industries. The return is calculated as the change in log price, day-to-day, as percent changes in closing price. Table 1 and Figure 1 provide a variety of descriptive statistics for returns. Except the negative mean daily return of finance index, other three indices have positive mean returns around 0.01% to 0.03%. Because of the different market size of indexes, the TAIEX daily return was less volatile than other 9.

(13) indexes according to the standard deviation. The four samples displayed excessive kurtosis, but exclusive of the finance, TAIEX, OTC and Electronic index skewed to the right.. [Insert Table 1 and Figure 1 here]. On the other way, for measuring the volatility spillover, we use Parkinson`s (1980) method to estimate the intraday variance of each index. For market i on day t we have. 政治大.  it2  0.361 ln( Pitmax )  ln( Pitmin ) . 立. 2. where ln( Pitmax ) is the maximum price in market i on day t and ln( Pitmin ) is the daily. ‧ 國. 學. minimum price. The  it2 is an estimator of the daily variance, so we convert it into. ‧. the annualized daily percent standard deviation is ˆ it2  100 365   it2 . In the same. sit. y. Nat. way, the summary statistics are described in Table 2 and Figure 2. From the. n. al. er. io. observation of mean and standard deviation, the volatility of TAIEX was the smallest.. i n U. v. Besides, they also performed excessive kurtosis but skewed to the left.. Ch. engchi. Last, we could see the past pattern on Figure 1 and Figure 2. In a rough way, the four indices follow the same trail in the past ten years. The return and volatility fluctuated significantly before 2003, but it decreased and remained the low level in the following years between 2004 and early 2007. After the sub-prime crisis, it rebounded to the high level in 2008. –. [Insert Table 2 and Figure 2 here]. 10.

(14) 4. Empirical study in Taiwan market. In practice, the comovement in the industry factors and the stock market is apparently observed in Taiwan`s stock trading activity, which we use four indices to represent. Academically, we do not use the traditional heteroskedastic model (Engle, 1990), otherwise we provide a variance decomposition measure to test the phenomenon. By decomposing the forecast error variance components for variable i coming from shocks to variable j, for all i and j, the spillover index is provided to. 政治大. explore the return and volatility interaction within variables.. 立. As a result, we characterize the spillover index as a two-way indicator. In order to. ‧ 國. 學. capture the source of the information transmission, we develop three definitions of spillover indices. Total spillover index is used to quantify the total spillover effect in. ‧. stock market and industries. Directional spillover index is to measure the. y. Nat. sit. contributions from other variables and the contributions to others. The net spillover is. n. al. er. io. the “to minus from” difference so that depicting the net influence on other variables.. i n U. v. In fact, many events historically affect Taiwan`s stock market through November. Ch. engchi. 1995 to April 2010 such as Asian financial crisis or dotcom bubble. It seems that a varying-parameter spillover analysis could apply over the entire sample. Hence we last introduce a dynamic study into the entire sample to assess the extent and nature of spillover variation over time. Specifically, the later context is separated into full-sample and rolling-sample analysis. We would calculate the total, directional and net spillovers in full-sample analysis in section 4.1. And then we examine the dynamics of spillover indices by rolling-sample method in sections 4.2.. 11.

(15) 4.1 Full- sample spillover analysis We begin to characterize daily return and volatility over the TAIEX, OTC, Electronics and Finance indices entire sample. Table 3 and Table 4 report various spillovers indices for the return and volatility respectively, which called Spillover Tables. The ijth entry in the table is the 10-day-ahead estimated contribution to the forecast error variance of index i (returns in Table 3 and volatilities in Table 5) coming from shock to index j. Hence the off-diagonal column sums are labeled. 政治大 the total spillover index appears in the lower right corner of the spillover table. It 立 directional to others and the row sums are labeled directional from others. In addition,. approximately is the off-diagonal column (or row sum) relative to the column sum. ‧ 國. 學. including diagonals (or row sum including diagonals), expressed as percentage. These. er. io. sit. y. Nat. spillover effect.. ‧. two spillover tables provide a “to and from” two-way decomposition of the total. n. Table 3 and Table 4 here] a[Insert iv l C n hengchi U. As we see the “Directional TO Others” row in Table 3, the gross directional return spillovers to others from the electronics are relatively large at 73.26 percent. And the “Directional FROM Others” column that gross directional return spillovers to others also shows the TAIEX is relatively large at 75.52%. Similarly, the volatility spillovers in Table 4 perform the same results as returns. The gross directional volatility spillovers to others from the electronics are relatively large at 62.34% and the gross directional volatility spillovers from others from the TAIEX are the greatest at 72.95%. Based on both kinds of spillovers, the evidence shows that the electronic industry has much more spillovers to others and TAIEX standing for Taiwan`s stock market has 12.

(16) enormous spillovers from others. The phenomenon represents the apparent electronic industry effect toward the stock market. The basic idea is that the industry containing the economic fundamentals information would influence the total stock market. According to Hong, Torous and Valkanov (20075)`s findings, the industry which is an indicator to forecast future economics could via gradual information diffusion influence the performance of stock market. In reality, the electronic industry is the core part of Taiwan`s economic activities. The largest directional spillovers form electronics to finance and other stock. 政治大 situation. Therefore based on the boudedly-rational-investor assumption (Shiller, 2000; 立 markets truly reflect the information transmission of investor`s prospect and economic. Sims, 2001), TAIEX receives the most directional spillovers from other industries`. ‧ 國. 學. information spread.. ‧. Finally, consider the right lower corner of the spillover tables. Distilling all the. sit. y. Nat. directional spillovers into a single spillover index, we find that almost 60% of forecast. io. er. error variance comes from spillovers both for returns at 63.93% and volatility at 58.18%. Hence the spillovers play an important role on the interaction of stock and. n. al. Ch. industries for both returns and volatility.. engchi. i n U. v. Statistically, we also conduct the significant test of the directional spillover 10-day-ahead forecasting. The nonparametric bootstrapping method (Efron, 1979) is applied on the accuracy of the sample statistics. Assuming the sample is an empirical distribution, each probability of data point is equal in our example. We repeatedly resample 1500 data points but eliminate the first 500 to avoid the sampling error from each replication. Finally, we simulate 1000 replications to calculate the 95% confidence interval and p-value for the directional spillovers. The statistical results are presented in Table 5 and Table 6. In the significant test for daily returns spillover in Table 5, the FROM and TO spillovers are all significant 13.

(17) at 1% level. For NET spillovers, except the finance is insignificant, TAIEX, OTC and electronics are significant. It reveals the directional spillovers are roughly significant excluded the obscure finance net spillover. In the same measure, the FROM and TO spillovers on daily volatilities in Table 6 are also significant at 1% level. However, in spite of TAIEX and OTC, the net spillovers for two industries electronics and finance are not significant. Obviously, in full-sample analysis, we have to process the net spillovers on electronic and finance carefully because of the offset effect of the TO and FROM spillovers so that the both are not statistically significant in Taiwan.. 政治大 [Insert Table 5 and Table 6 here] 立 ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 14. i n U. v.

(18) 4.2 Rolling-sample spillover analysis Indeed, many events happened during our sample period 1995-2010. Since the globalization accelerates the capital mobilization, Taiwan stock market has sustained the turbulence and innovation from the world market, even from the domestic evolution and policies. Under the background of irregular jointly influence during the periods, the fixed-parameter model would be weakly applied on the sample. Although we obtain very useful description of average behavior from the preceding full-sample. 政治大 movements in Taiwan stock. Therefore, to do a more comprehensive survey, we use 立. spillovers, unfortunately, it may invisibly miss critical cyclical and specific. the 100-day rolling method over our samples. We present the variation of spillover. ‧ 國. 學. index on the plots. In order to succinctly explaining the dynamics in plots , we would. ‧. mainly focus on the volatility spillovers of the directional and net spillovers.. er. io. sit. y. Nat. [Insert Figure 3 here]. al. n. v i n C h results of volatility First, we examine rolling-sample and return total spillovers engchi U. in Figure 3. The return usually fluctuates between sixty and sixty-five percent and the volatility is at the fifty-five and sixty percent range. The high level of spillovers among the industries and the stock market implies the intense relation on each other. By receiving the same signal from the market, the return and volatility both display the similar trend pattern. They respond strongly to the economic events and recover to the stable state in the regular range. We remark the economic events which could influence Taiwan directly below to explain the increasing fluctuation of return and volatility in the events: 15.

(19) (1). In late 1997, the currency of Taiwan`s neighborhood Thailand devaluated dramatically within a few days, then the currency devaluation panic spread to other major economics such as Hong Kong, South Korea etc..Inevitably, East Asian financial crisis cause TAIEX fell immensely 30% from original 10,000 points.. (2). In 2000-2001, mostly due to the political factors, the president election expanded the investment risk in Taiwan. In addition, the dotcom bubble in NASDAQ brought about the downturn of Taiwan`s electronic industry which is. 政治大 With the weave of foreign direct investment and merger into Taiwan and the 立. a mainstream in Taiwan economics. (3). economic propensity of neighboring Asian stock markets, Taiwan`s stock. ‧ 國. 學. market rose abruptly in 2006.. The global financial market disaster associated with the USA subprime. ‧. (4). sit. y. Nat. mortgage default in late 2007 to 2008 reduced Taiwan`s stock market value. io. er. approximately half.. Given above-mentioned events, they produce large return and volatility spillovers. al. n. v i n C and among TAIEX, OTC, electronics Except the stock market consolidation h efinance. ngchi U. in 2004-2005, each economic or political event fostered total return or volatility spillovers beyond their average level and provides a reason for the high spillovers.. The extreme change of the economic fundamental foster the spillover effect among industries and stock market, under such a situation, the information spread shortly in days even hours therefore enlarging the scale of total spillovers in separated portfolios. So far we have discussed the total spillovers, we next detect the directional volatility spillovers in following Figures 4 and 5. Because the return and volatility 16.

(20) spillovers display the likely moving pattern, to make the plots and analysis plainer, we choose the volatility spillovers on behalf in the rolling sample.. [Insert Figure 4 and Figure 5 here]. We put the “FROM” volatility spillovers plot in Figure 4. Compared to the dynamic total spillovers, the FROM spillovers vary greatly over time. The FROM spillovers of finance provides the largest sway, and OTC and electronics are in the. 政治大 spillovers is smoother regardless of the economic impact, while other three indices 立. next place, last TAIEX performs the smallest variation. On average, TAIEX`s FROM. respond fiercely to the events. The reason is that the industry indices as electronics. ‧ 國. 學. and finance are sensitive to the boom and recession of the economics, consequently,. ‧. reflecting on the vibration of the volatility spillovers.. sit. y. Nat. In Figure 5 we presented the directional volatility spillovers to our four indexes.. io. er. Overall, the total variation of TO spillovers is apparently smaller than FROM spillovers. For TAIEX, the TO spillovers is less volatile than other TO spillovers as. al. n. v i n C h we could observe well as its FROM spillovers. Instead, that OTC and finance are engchi U more sensitive to the events, especially finance, which is the devastated industry in. 1997 Asian financial crisis and 2007 subprime mortgage storm but a prosperous invested target in 2006. Interestingly, there is a clear upward trend of electronics, manifesting the increasingly decisive role the electronic industry plays in Taiwan market. The good or bad news diffused from the electronic sectors gets more and more valuable on the prediction and explanation in market portfolios. In the summary of the directional spillovers, we finally discuss the net spillovers in figure 6. Net spillovers from TAIEX appear consistently negative and large, acting as a receiver taking over other markets` impact. The net spillovers 17.

(21) declined to -50 percent at Asian financial crisis, -20 percent at Dotcom bubble and -40 percent after mortgage collapse. Net spillovers of other three indices are not easily indentified as the inflow or outflow influence on others. We could observe the steeply rising outflow spillovers from OTC, electronics and finance in above three crisis times. In 1997, the dramatic depreciation of New Taiwan Dollar has an positive impact on the exporting-oriented electronic industry which then produced high positive net spillovers as well as the panic finance industry bearing the shock from other foreign capital market. Following the technology disaster in 2000, electronics. 政治大 The same results again appear in late 2007, these three indices changed upwards, but 立. and OTC show suddenly positive net spillovers while the finance also climbed fast.. noticeably finance provided large net spillovers thereafter in following two years.. ‧ 國. 學. While the above events happened, the increasing spillover effect from industries. ‧. and falling spillover from the stock market proved the importance of industry factors. sit. y. Nat. in explaining the stock market movement. The big financial event brings the strong. io. er. and transient information passing through stock market from the industries which would blow the warning horns in early stages of economic crisis.. n. al. Ch. engchi. 18. i n U. v.

(22) 5. Robustness. In the last section, we perform some robustness test on our basic analysis. First, the whole sample would be divided into subsamples according to the time-varying transition of Taiwan`s industry weight, investigating the factor whether determined the spillover effects. Next, we introduce the weekly and monthly data replaced the original daily data to check the consistency toward the different time series.. 立. 5.1 Subperiods. 政治大. ‧ 國. 學. From the textile and food industries in 1950s, the plastics, concrete and construction industries in 1960s to 1970s, and the electronics and finance of 1980s. ‧. until now, Taiwan`s main economic activities has transited in past generations owing. y. Nat. io. sit. to the policies or economic cycles. Hence, the weight of individual industry has. n. al. er. changed historically according to the market value of in-listed companies the industry. i n U. v. comprised. In the context we mentioned before, we calculate the spillovers of entire. Ch. engchi. sample in 1995-2010, however, ignoring the weight change of industries toward the stock market. In the study, we have to ensure the spillover effect from industries to stock market or vice versa not due to the compiling of indices. Actually, the spillover effect is based on the natural comovement among the industry and stocks rather than the artifact method. In Figure 7, we present the transition of industry weight relative to TAIEX over the sample period. The finance had more weight than the electronics around 15% in late 1990s, but after the Asian financial crisis, the finance stock market value gradually withered and electronics just obtain the strength from the global technology 19.

(23) propensity so that increasing enormously to achieve half weight of stock market. The flux and reflux of these two industries may affect our spillover results, in that case we use 1998 which is a watershed of weight as the midpoint to produce two subsamples “1995-1998” and “1999-2010”.. [Insert Figure 7 here]. We employ the full-sample method in the subsamples research and present the. 政治大 spillover summaries . At first, we examine the total spillover effect within the 立. return and volatility spillovers in Table 7 through Table 10 to conclude the daily 3. different periods in the lower right corner. For the returns, the total spillovers. ‧ 國. 學. increased 10 percent over the periods, in a likely manner, the total spillover of the. ‧. volatility jumped from 45 percent to 60 percent. The increment of spillovers. sit. y. Nat. throughout the years reflects the closer relationship between industries and stocks as a. io. al. n. movement.. er. result of the healthier capital market and the less strict government rule of funding. Ch. engchi. i n U. v. [Insert Table 7-10 here]. Next comparing the directional spillovers across the periods, for returns, we could observe that TAIEX and electronics have higher FROM and TO spillovers about 70 percent, respectively, which is also consistent with the whole sample results before. Looking into another table, the volatility spillovers perform the same outcomes showing the largest TO and FROM spillovers still provided by electronics and TAIEX. Aggregating the consistent directional spillovers results across 3. We could make an reference to Table 3 and 4 to compare our sub-period outcome. 20.

(24) subperiods without the influence of the weight change, we could infer that the stock market indices comprising sample industries is not the reason for the mutual spillover effect. So it is not likely that the spillover exists among industries and stock market because the industries are parts of the market portfolio, probably due to the economic fundamental connection or funding liquidity.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 21. i n U. v.

(25) 5.2 Weekly and monthly data Instead of the original daily data, we switch our framework toward the longer term data- weekly and monthly returns and volatilities so as to find whether the consistency of spillover effects exists in different times. Either weekly or monthly data we convert to are spanned from November, 1995 to April, 2010 as the same as the forecited sample period and accumulated 750 and 174 samples respectively. We directly report the significant test by the bootstrapping method in directional. 政治大 directional spillovers to daily data but also check the statistical accuracy. 立. spillovers in Table 11 through Table 14, in that we could not only compare the. In this section, we separately specify the return and volatility spillover results in. ‧. ‧ 國. 學. io. sit. y. Nat. [Insert Table 11-14 here]. er. sequence.. We first look over the weekly and monthly return directional spillovers in Table. al. n. v i n 11 and 12. Recall that in dailyCdata the FROM spillovers h e n g c h i U are concentrated on the 4. TAIEX and so does the electronic industry which produces most TO spillovers in the four indices. The FROM spillovers in weeks and months are about seventies percent approaching the daily return spillover 75%. And the electronics also perform the highest TO spillovers at 73% in monthly samples, however, in weekly returns, the OTC index displays the largest TO spillovers. Although it is not consist with prior results, the electronic still has approximately 70 percent keeping in the high level of influence. Net spillovers show results well matched with the daily data. The enormous negative net spillovers emerge from TAIEX and electronics appear the positive 4. We can refer to Table 5 and Table 6 to get more information in daily data. 22.

(26) spillovers at the 5% significance level. Finally, we are still uncertain about the direction of net spillovers of finance owing to the inconsistency among different time series data. Before the discussion of volatility spillovers, we employ specific methods in order to process the weekly and monthly variance calculation. For weekly volatility, we follow the work from Garman and Klass (1980) and Alizadeh et al (2002) to obtain the weekly return variance:  2  0.511( Ht  Lt )2  0.019 (Ct  Ot )( Ht  Lt  2Ot )  2( Ht  Ot )(Lt  Ot )  0.383(Ct  Ot )2. 政治大 open and C is the Friday立 close (all in natural logarithms). On the other way, we. where H is the Monday-Friday high, L is the Monday-Friday low, O is the Monday. ‧ 國. 學. calculate the monthly volatility spillovers as the variance of each month`s daily returns. Given above characterization of samples, we report the results in Table 13 and. ‧. Table 14.. sit. y. Nat. Have in mind that TAIEX and electronics has the most FROM and TO spillovers,. n. al. er. io. respectively, and TAIEX takes larger net spillover from others while electronics and. i n U. v. finance have obscure directions in daily volatility spillovers. As with the directional. Ch. engchi. volatility spillover FROM others, TAIEX is influenced by others mostly no matter in weeks or months. Nevertheless, except the electronics maintains high spillovers to others in weekly data, it reveals an inconsistency in monthly data which shows the finance as the largest spillovers outputted to others. Last, TAIEX still plays the role as a spillovers receiver from the negative net spillovers. In contrast, the direction of net spillovers of others could not be determined because of the insignificance of OTC and the contradiction happened on electronics and finance in both time series. Combining all the robustness results in different times and returns or volatility spillovers in this section, we could conclude that TAIEX undoubtedly receive most 23.

(27) spillovers from others even surpassing the TO spillovers it produces. In the long run spanned from weeks and months, we could not absolutely make sure which FROM and TO spillovers industries provide is larger so that recognizing the electronic or finance as a net spillovers outputer or an inputer.. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 24. i n U. v.

(28) 6. Conclusion. In this paper, we have provided both total and directional spillover measures that are irrelevant to the variable ordering used for forecast error variance decompositions. We applied this method on the returns and volatilities of comovement among Taiwan stock market and industries in static and dynamic analysis. The total spillovers account for a large part of the stock market and industry price motion. Overall, observed from the FROM and net spillovers, the stock market. 政治大 there`s no obvious directions 立for industries to influence or to be influenced. Finally,. performs as an undertaker from the information diffused from the industries. However,. ‧ 國. 學. we simulate the time-varying spillover index over last fifteen years, discovering the fierce swing of industry spillovers and smoother motion of stock market. The results. ‧. reveal the volatility transmission from the industries to stock market during the recent. sit. y. Nat. crisis or tremendous change in economics. Investors ought to be cautious about the. n. al. er. io. information like economical news or regulation diffused from industries especially. i n U. v. electronics, so then actively adjusting the weight of stock market portfolio. If there is. Ch. engchi. negative news such as the withered consumption producing from industries, investors should quickly decrease the spot stock market or futures market portfolios to avoid the following spillover effect shock.. 25.

(29) References Alizadeh, S., Brandt, M.W. and Diebold, F.X. (2002). “Range-based estimation of stochastic volatility models,” Journal of Finance, vol. 57(3):, 1047–92. Bollerslev, Tim, (1986), “Generalized autoregressive conditional heteroskedasticity,” Journal of Econometrics 31, 307-327. Cavaglia, Stefano, Dimitris Melas, and George Tsouderos. (2000). “Cross-Industry and Cross-Country International Equity Diversification”. The Journal of Investing. Vol. 9, no.1: 65-71 Diebold, F.X. and Yilmaz, K. (2009), "Measuring Financial Asset Return and Volatility Spillovers, With Application to Global Equity Markets," Economic Journal, 119: 158-171. Diebold, F.X. and Yilmaz, K. (2010), "Better to Give than to Receive: Predictive Directional Measurement of Volatility Spillovers," International Journal of Forecasting, forthcoming. (With discussion.) Edwards, S. and Susmel, R., (2001).” Volatility dependence and contagion in emerging equity markets.” Journal of Development Economics, 66: 505-532. Edwards, S. and Susmel, R.,( 2003).” Interest-rate volatility in emerging markets.”. 立. 政治大. ‧. ‧ 國. 學. The Review of Economics and Statistics, 85 :328-348.. Nat. y. sit. n. al. er. io. Engle, Robert E., (1982).” Autoregressive conditional heteroskedasticity with estimates of the variance of United Kingdom inflation,” Econometrica 50: 987-1007. Engle, R.F., Ito, T., Lin,W., (1990).” Meteor showers or heat waves? Heteroskedastic intra-daily volatility in the foreign exchange market.” Econometrica 58: 525–542. Eun, C.S., Shim, S. (1989), “International Transmission of Stock Market Movements,” Journal of Financial & Quantitative Analysis, 24: 241-256. Forbes, K., Rigobon, R. (2002), “No Contagion, only Interdependence: Measuring Stock Market Comovements,” Journal of Finance, 57: 2223-2261.. Ch. engchi. i n U. v. Garman, M.B. and Klass, M.J. (1980). “On the estimation of security price volatilities from historical data,” Journal of Business, vol. 53(1): 67–78. Hamao, Y., Masulis, R.W., Ng, V. (1990), “Correlations in Price Changes and Volatility across International Markets,” Review of Financial Studies, 3: 281-307. Hamilton, J.D., (1989). “A new approach to the economic analysis of nonstationary time series and the business cycle.” Econometrica 57:357–384 Hamilton, J.D. and Susmel, R., (1994). “Autoregressive conditional heteroskedasticity 26.

(30) and changes in regime.” Journal of Econometrics, 64 :307-333. Harvey, C.R. (1995), “Predictable Risk and Returns in Emerging Markets,” Review of Financial Studies, 8: 773-816. Heston, Steven L. and K. Geert Rouwenhorst. (1994). “Does Industrial Structure Explain the Benefits of Industrial Diversification?” Journal of Financial Economics, Vol. 36, no.1: 3 –27. Hong, H., Stein, J.,( 1999).” A unified theory of underreaction, momentum trading and overreaction in asset markets.” Journal of Finance 54 :2143–2184. Hong, H. and Torous, W. and Valkanov, R. (2007), “Do Industries Lead Stock Markets?” Journal of Financial Economics 83:367-396. Karolyi, G.A. (1995). “A Multivariate GARCH Model of International Transmissions of Stock Returns and Volatility: The case of the United States and Canada,” Journal of Business and Economic Statistics, 13 :11-25. King, Mervyn A and Wadhwani, Sushil, (1990). "Transmission of Volatility between Stock Markets," Review of Financial Studies, Oxford University Press for Society for Financial Studies, vol. 3(1) :5-33. King, M., Sentana, E. and Wadhwani, S. (1994), “Volatility and Links Between National Stock Markets,” Econometrica, 62 : 901-933. Koop, G., Pesaran, M.H., and Potter, S.M. (1996), “Impulse Response Analysis in Non-Linear Multivariate Models,” Journal of Econometrics, 74: 119–147.. 立. 政治大. ‧. ‧ 國. 學. n. al. er. io. sit. y. Nat. Lessard, Donald. (1974). “World, National, and Industry Factors in Equity Returns”. Journal of Finance. Vol. 29, no. 3: 379-391. Morana, C., Beltratti, A. (2006), “Comovements in International Stock Markets,” Journal of International Financial Markets, Institutions and Money, forthcoming. Parkinson, M. (1980), “The Extreme Value Method for Estimating the Variance of the Rate of Return,” Journal of Business, 53: 61-65. Pesaran, M.H. and Shin, Y. (1998), “Generalized Impulse Response Analysis in Linear Multivariate Models,” Economics Letters, 58: 17-29. Richards, A.J. (1995), “Comovements in National Stock Market Returns: Evidence of. Ch. engchi. i n U. v. Predictability, but not Cointegration,” Journal of Monetary Economics, 36: 631-654. Shiller, R., (2000).” Irrational Exuberance.” Broadway Books, New York. Sims, C., (2001).” Rational Inattention.” Princeton University Working Paper.. 27.

(31) Table 1 Descriptive Statistics of Taiwan Four Indices Daily Returns Mean Std. Dev. Skewness Kurtosis Maximum Minimum Median. TAIEX. OTC. Electronics. Finance. 0.000125 0.015570 -0.152000 4.934000 0.065250 -0.069760 0.000284. 0.000090 0.017600 -0.132600 4.220000 0.064390 -0.069680 0.000207. 0.000296 0.018560 -0.139300 4.305000 0.065160 -0.069690 0.000557. -0.000088 0.018720 0.080760 4.878000 0.066390 -0.070850 -0.000672. Notes: The period is from Nov. 3, 1995 to Apr. 30, 2010 and the sample size is 3743. The MEAN, STANDERD DEVIATION, SKEWNESS, KURTOSIS, MAXIMUM, MINIMUM. 政治大. and MEDIAN are measured in day-to-day closing price percent change. The TAIEX ,. 立. Electronics Index and Finance Index is compiled by TWSE on listed companies according to the weight of individual market value. OTC Index is edited by GTSM on the. ‧. ‧ 國. 學. over-the-counter trading companies according to the weight of individual market value.. Electronics. 18.4800 10.7500 1.7690 8.3270 109.1000 1.6780 15.9500. 21.6700 13.2800 1.4860 5.7080 97.2100 0.9716 18.1100. 22.3400 12.6000 1.6510 7.5960 125.2000 1.5450 19.3100. Ch. engchi. er. n. al. y. OTC. sit. TAIEX. io. Mean Std. Dev. Skewness Kurtosis Maximum Minimum Median. Nat. Table 2 Descriptive Statistics of Taiwan Four Indexes Daily Volatilities. i n U. v. Finance. 23.0600 13.6200 1.5560 6.4250 111.4000 1.4100 19.8600. Notes: The period is from Nov. 3, 1995 to Apr. 30, 2010 and total observations are 3743. The daily variance is estimated by Parkinson`s method with daily most high and low price.. 28.

(32) Table 3 Spillover Table, Daily Returns TAIEX. OTC. Electronics. Finance. Directional FROM Others. TAIEX OTC Electronics Finance Directional. 24.48 20.27 23.23 21.54 65.05. 21.65 35.71 22.66 16.23 60.54. 29.37 26.95 38.83 16.95 73.26. 24.50 17.08 15.29 45.28 56.85. 75.52 64.29 61.17 54.72 255.71. TO Others Directional Including Own. 89.533. 96.25. 112.09. 102.13. Spillover index. 立. 政治大. 63.93%. Note: The underlying variance decomposition is based upon a daily generalized VAR of order 1. The (i,j)-th value is the estimated contribution to the variance of the 10-day-ahead stock return. ‧ 國. 學. forecast error of index I coming from shock to stock returns of index j. The data are defined as. ‧. io. TAIEX. y. Electronics Finance. n. al. OTC. sit. Nat. Table 4 Spillover Table, Daily Volatilities. er. Table 1.. TAIEX OTC Electronics Finance Directional TO Others. 27.05 17.30 22.83 18.98. 21.93 47.87 21.67 15.32. Ch. 27.48 20.43 41.11 14.43. 59.11. 58.92. Directional Including Own. 86.16. 106.79. i n U. v. Directional FROM Others. 23.54 14.40 14.39 51.27. 72.95 52.13 58.89 48.73. 62.34. 52.33. 232.70. 103.45. 103.60. Spillover index. engchi. 58.18%. Note: The underlying variance decomposition is based upon a daily generalized VAR of order 1. The (i,j)-th value is the estimated contribution to the variance of the 10-day-ahead stock return volatility forecast error of index i coming from shock to stock return volatility of index j. The data are defined as Table 2.. 29.

(33) Table 5 Significant Test for Directional Spillover, Daily Returns Spillover FROM Others. Lower CI. Upper CI. p-value. 75.52 64.293 61.174 54.722. 75.29 63.111 60.024 53.58. 75.734 65.462 62.007 55.464. 0*** 0*** 0*** 0***. Spillover TO Others. Lower CI. Upper CI. p-value. TAIEX OTC. 65.053 60.539. 64.551 59.072. 65.902 62.291. 0*** 0***. Electronics Finance. 73.264 56.853. 71.81 52.805. 75.077 59.337. 0*** 0***. Net Spillover. Upper CI. p-value. 12.09 2.1308. -9.3996 -1.538 15.043 5.1045. 0*** 0*** 0*** 0.125. TAIEX OTC Electronics Finance. ‧ 國. 10.019 -1.7012. 學. TAIEX OTC Electronics Finance. 治Lower CI 政 -10.467 -11.01 大立 -3.7536 -6.1043. Note:*** Significant at 1% level, ** Significant at 5% level, *Significant at 10% level.. ‧. a72.95 l C 52.13 h. 71.83 48.74 56.82 45.18. n. 58.89 48.73. engchi U. sit. Lower CI. Upper CI. er. Spillover FROM Others. io. TAIEX OTC Electronics Finance. y. Nat. Table 6 Significant Test for Directional Spillover, Daily Volatilities. v ni. p-value. 73.93 55.62 61.12 52.06. 0*** 0*** 0*** 0***. Spillover TO Others. Lower CI. Upper CI. p-value. TAIEX OTC. 59.11 58.92. 56.44 54.36. 61.78 63.61. 0*** 0***. Electronics Finance. 62.34 52.33. 57.7 47.25. 67.02 58.21. 0*** 0***. Net Spillover. Lower CI. Upper CI. p-value. -13.84 6.792 3.451 3.598. -16.92 0.8625 -2.483 -3.102. -10.8 13.16 9.048 10.84. 0*** 0.017** 0.136 0.165. TAIEX OTC Electronics Finance. Note:*** Significant at 1% level, ** Significant at 5% level, *Significant at 10% level. 30.

(34) Table 7 Spillover Table, Daily Returns 1995-1998 TAIEX TAIEX OTC Electronics Finance. Electronics. Finance. Directional FROM Others. 23.84 14.84 18.61 22.73. 20.06 47.7 18.65 17.34. 31.94 23.82 54.26 13.7. 24.16 13.63 8.48 46.24. 76.16 52.3 45.74 53.76. 56.18. 56.05. 69.46. 46.27. 228. 80.02. 103.7. 立. 治 92.51 政 123.7 大. Spillover index 56.99%. 學. ‧ 國. Directional TO Others Directional Including Own. OTC. Table 8. TAIEX. Electronics Finance. Directional. y. Nat. 28.79 27.67 34.52 17.74. 67.47. 61.58. 92.1. 94.14. n. al. Ch. 24.55 17.99 17.45 45.06. v. 75.36 67.44 65.48 54.94. 74.19. 59.99. 263.2. 108.7. 105. Spillover index. engchi. er. 22.02 32.56 23.61 15.95. sit. FROM Others. 24.64 21.78 24.43 21.26. io. TAIEX OTC Electronics Finance Directional TO Others Directional Including Own. OTC. ‧. Spillover Table, Daily Returns 1999-2010. i n U. 65.81%. 31.

(35) Table 9 Spillover Table, Daily Volatility 1995-1998 TAIEX OTC Electronics Finance TAIEX OTC Electronics Finance Directional TO Others Directional Including Own. Directional FROM Others. 31.88 9.72 16.9 20.17. 13.95 68.7 11.17 10.5. 30.68 14.91 62.04 14.86. 23.49 6.671 9.892 54.47. 68.12 31.3 37.96 45.53. 46.79. 35.62. 60.44. 40.05. 182.9. 78.67. 104.3. 立. 政 122.5治 94.52 大. Spillover index 45.73%. ‧ 國. 學. TAIEX. Nat. Electronics Finance. Directional. 27.73 22.18 35.3 15.31. 61.02. 65.17. 86.67. 108.7. n. al. Ch. 22.77 15.65 15.62 50.1. v. 74.35 56.5 64.7 49.9. 65.21. 54.04. 245.4. 100.5. 104.1. Spillover index. engchi. er. 23.85 43.5 24.98 16.34. sit. y. FROM Others. 25.65 18.67 24.1 18.25. io. TAIEX OTC Electronics Finance Directional TO Others Directional Including Own. OTC. ‧. Table 10 Spillover Table, Daily Volatility 1999-2010. i n U. 61.36%. 32.

(36) Table 11 Significant Test for Directional Spillover, Weekly Returns Spillover FROM Others TAIEX OTC Electronics Finance. 77.01 59.78 61.38 55.65. Lower CI Upper CI 76.25 56.86 58.16 51.15. Spillover TO Others. 77.83 62.47 64.2 59.92. Lower CI Upper CI. p-value 0*** 0*** 0*** 0*** p-value. TAIEX OTC. 60.52 75.74. 57.65 70.13. 63.32 81.87. 0*** 0***. Electronics Finance. 69.89 47.67. 63.41 40.14. 75.98 55.24. 0*** 0***. Net Spillover. Upper CI. p-value. 8.508 -7.976. -13.09 23.81 16.46 0.4491. 0*** 0*** 0.019** 0.037**. ‧. ‧ 國. 0.5239 -16.63. 學. TAIEX OTC Electronics Finance. 治Lower CI 政 -16.49 -20.06 大立 15.96 9.257. 75.24 62.81 59.9 52.61. n. al. Ch. engchi. Upper CI. p-value. 75.82 65.64 62.35 56.57. 0*** 0*** 0*** 0***. er. 75.52 64.29 61.16 54.69. io. Lower CI. sit. Nat. TAIEX OTC Electronics Finance. Spillover FROM Others. y. Table 12 Significant Test for Directional Spillover, Monthly Returns. i n U. v. Spillover TO Others. Lower CI. Upper CI. p-value. TAIEX OTC Electronics. 65.04 60.5 73.26. 64.02 58.47 70.83. 65.97 62.63 75.39. 0*** 0*** 0***. Finance. 56.87. 53.65. 60.1. 0***. Net Spillover. Lower CI. Upper CI. p-value. -10.47 -3.799 12.1 2.177. -11.71 -6.391 9.066 -1.428. -9.322 -1.272 14.86 5.974. 0*** 0*** 0*** 0.112. TAIEX OTC Electronics Finance. 33.

(37) Table 13 Significant Test for Directional Spillover, Weekly Volatilities Spillover FROM Others. Lower CI. Upper CI. p-value. 78.39 54.92 52.99 52.82. 76.2 43.72 45.61 41.26. 81.39 64.42 60.25 63.23. 0*** 0*** 0*** 0***. Spillover TO Others. Lower CI. Upper CI. p-value. TAIEX OTC. 48.28 64.62. 38.47 50.22. 57.27 81.05. 0*** 0***. Electronics Finance. 76.09 50.13. 56.64 36.82. 94.55 68. 0*** 0***. Net Spillover. Upper CI. p-value. 23.1 -2.693. -19.27 31.86 46.59 22. 0*** 0.201 0.026** 0.417. TAIEX OTC Electronics Finance. ‧. ‧ 國. -0.5974 -22.77. 學. TAIEX OTC Electronics Finance. 治Lower CI 政 -30.11 -42.54 大立9.702 -10.26. 79.99 64.25 63.94 46.63. n. al. Ch. 77.09 55.38 56.34 37.28. engchi. Spillover TO Others. sit. io. TAIEX OTC Electronics Finance. Lower CI Upper CI 82.71 71.88 72.81 58.41. er. Nat. Spillover FROM Others. y. Table 14 Significant Test for Directional Spillover, Monthly Volatilities. i n U. v. Lower CI Upper CI. p-value 0*** 0*** 0*** 0*** p-value. TAIEX OTC Electronics. 49.56 58.49 62.72. 42.31 40.85 44.44. 57.19 74.3 87.11. 0*** 0*** 0***. Finance. 84.03. 55.91. 115.6. 0***. Net Spillover TAIEX OTC Electronics Finance. Lower CI Upper CI. -30.43 -5.755 -1.218 37.4. -39.96 -26.77 -26.62 -2.085. 34. -20.3 13.25 28.55 75.87. p-value 0*** 0.264 0.457 0.032**.

(38) TAIEX. OTC 0.08 0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08 2009. 2007. 2005. 2003. Electronic 0.08 0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08. 2009. 2005. 2003. 2001. Nat. io. sit. Figure 1 Daily Returns of Taiwan Four Indices. n. al. er. 1999. ‧ 國. 1997. ‧. 1995. 立. 2007. 政治大. 學. 0.08 0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08. Finance. y. 2001. 1999. 1997. 1995. 0.08 0.06 0.04 0.02 0 -0.02 -0.04 -0.06 -0.08. Ch. engchi. 35. i n U. v.

(39) 120 100. 政治大 80 60 20. 2001. 1999. 1997. 1995. 0. 2003. 40. ‧. ‧ 國. 立. Figure 2 Daily Volatilities of Taiwan Four Indices. n. al. er. io. sit. y. Nat. \. 2009. Finance. 學. 140 120 100 80 60 40 20 0. 2009. 1995. Electronic. 2007. 0. 2007. 0. 2005. 20. 2005. 20. 2003. 40. 2001. 40. 2009. 60. 2007. 60. 2005. 80. 2003. 80. 2001. 100. 1999. 100. 1997. 120. 1995. 120. 1999. OTC. 1997. TAIEX. Ch. engchi. 36. i n U. v.

(40) Volatility. Return. 75 70 65 60. 55 50 45 40 35 30 25 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010. 立. 政治大 Figure 3. Spillover Plot, Taiwan Four Indices Returns and Volatility. ‧ 國. 學. Notes:We plot moving return and volatility Spillover indices, defined as the sum of all variance decomposition “directional FROM others” or “directional TO others”, estimated using 100-day rolling. ‧. window.The horizonal axis is the ending date of window... n. er. io. sit. y. Nat. al. Ch. engchi. 37. i n U. v.

(41) TAIEX. OTC. 90. 120. 80 100. 70 60. 80. 50. 60. 40 30. 40. 20. 20. 10. 2008. 2010 2010. 2004. sit 2004. v. 2000. engchi. i n U 1998. 1996. Ch. 0. 2010. al. er. 20 2008. 2004. 2002. 2000. 1998. 40. n. 1996. y. 60. io. 0. 2008. 80. 2006. 20. 2006. 100. 2002. ‧ 國. 120. Nat. 40. 2002. 140. ‧. 60. Finance. 學. 80. 2006. 立. 100. 2000. 1996. 政治大. Electronic 120. 1998. 0. 0. Figure 4 Directional Volatility Spillovers, From Taiwan Four Indices Notes: The moving directional spillovers is calculated from other influence, defined as the variance decomposition “Directional FROM others”, estimated using 100- day rolling windows. The horizonal axis is the ending date of window.. 38.

(42) TAIEX. OTC. 90. 80. 80. 70. 70. 60. 60. 50. 50. 40. 40. 30. 30. 20. 20. 10. 10. 2008. 2010 2010. 2004. y. 0 1996. 2010. 2008. Ch. engchi. i n U. v. 2004. al. 2006. 2004. 2002. 2000. 1998. 1996. sit. 10. n. 0. 20. 2002. io. 10. 30. er. 20. 2008. 40. Nat. 30. 2002. 50. 2000. 40. 60. ‧. 50. 70. 1998. 60. 2000. 1996. 2010. 2008. 80. ‧ 國. 70. Finance. 學. 80. 2006. 立. 政治大. 2006. 90. 2006. 2004. 2002. 2000. 1998. 1996. Electronic. 1998. 0. 0. Figure 5 Directional Volatility Spillovers, TO Taiwan Four Indices Notes: The moving directional spillovers is calculated the influence to others, defined as the variance decomposition “Directional TO others”, estimated using 100- day rolling windows. The horizonal axis is the ending date of window.. 39.

(43) -60. Electronic. 100. 立. 60 40. Finance. 政治大. 80. 2010. -60. 2010. -40 2008. -50. 2008. -20. 2006. -40. 2006. 0. 2004. -30. 2002. 20. 1996. -20. 2010. 40. 2008. -10. 2006. 60. 2004. 0. 2002. 80. 2000. 10. 1998. 100. 1996. 20. 2000. OTC. 1998. TAIEX. 80. Ch. i n U. v. 2004. y e2000 rs 2002 i t. 1998. 2010. al. 1996. -60 2008. 2000. -40. n. 1998. io. 1996. -80. -20. Nat. -60. 0. 2006. -40. 20. ‧. -20. 40. 2004. 0. 2002. 20. 學. ‧ 國. 60. Figure 6 Net Directional Volatility Spillovers, Taiwan Four Indices. engchi. Notes: The moving net directional spillovers is calculated the net influence to or fromothers, defined as the variance decomposition “Directional TO others minus directional from others”, estimated using 100day rolling windows. The horizonal axis is the ending date of window.. 40.

(44) Electronic Weight. Finance Weight. 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1. 1999 2000 2001 2001 2002 2003 2003 2004 2005 2005 2006 2007 2007 2008 2009 2009. 1999. 1998. 1997. 1997. 1996. 1995. 0. Figure 7 The transition of Industry Weight Relative to TAIEX. 立. 政治大. Notes: The weight is based on the daily market value of electronics and finance relative to TAIEX in1995-2010. The Source is from TEJ database and edited by author.. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 41. i n U. v.

(45) 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 42. i n U. v.

(46)