Price Transmission Effect between GDRs and Their Underlying Stock

Price Transmission Effect between GDRs and Their

Underlying Stocks—Evidence from Taiwan

SHEN-YUAN CHEN∗

Department of Finance, Ming Chuan University E-mail: sychen@mcu.edu.tw

LI-CHUAN CHOU

Department of Finance, Ming Chuan University

CHAU-CHEN YANG

Department of Finance, National Taiwan University

Abstract. In this paper we examine the price transmission effect between ADRs or GDRs and their respective underlying stocks. This linkage is investigated for Granger causality using difference form and VECM. Results reveal unidirectional causality from Taiwan’s capital market to the foreign market. This asymmetry suggests the domestic market plays a dominant role in price transmission relative to the foreign market. Besides, the prices of both markets will make adjustment to establish a long run cointegrated equilibrium. An additional finding is that both the premium and net buy have significant impacts on international price transmission for over twenty percent samples. Empirical outcomes also provide the evidence that our model is quite robust.

Key words: price transmission, ADRs, GDRs, premium, net buy

JEL Classification: G15, F21, F23, C22

1. Introduction

Among the emerging equity markets, Taiwan’s capital market is an increasingly important one for global institutional investors due to the government’s incessant revolution and liberalization policy in past decade. For example, in 1991, foreign institutional investors were allowed to directly invest in Taiwan stock market and from September 1996, Taiwan market was included in its indices by the Morgan Stanley Capital International Inc. (MSCI). As Taiwan capital market continuously deregulated, foreign investors are getting more active to this emerging market.

At the same time, as an important Original Equipment Manufacturer for worldwide famous enterprises in recent years, Taiwan companies, especially for high-tech industries, are attracting more attention from the global investors. Rapid growth in competitive ability has engendered the result that a large number of Taiwan firms have their stocks cross-listed ∗_{Address correspondence to: 250, Chung Shan N. Rd. Sec. 5, Taipei, Taiwan. Tel.: (886) 2-28824564-2390, Fax:}

on international exchanges successfully. The stock price linkage between Taiwan market and foreign market has become an important issue for local and foreign institutional investors because of the price interaction and arbitrage opportunities provided by dually listing.

The most commonly used vehicles of dual listing by Taiwan companies are American Depositary Receipts (ADRs) and Global Depositary Receipts (GDRs). The possible advan-tages for such cross listing are promoting a firm’s reputation in large capital markets, the availability of capital, lower capital costs and elimination of investment barriers such as do-mestic accounting and tax practices (see Karolyi (1998) survey on why and how companies list abroad).

DRs can be created in one of the two ways: sponsored DR and unsponsored DR. In a sponsored DR, the underlying corporation pays a fee to the depositary institution to cover the cost of DR program. By regulation, the underlying corporation must provide periodic financial reports to the holders of DRs. A sponsored DR is often issued by a public company to seek to have its stock traded in foreign country and to raise capital from a foreign market. In contrast to sponsored DR, an unsponsored DR is issued by one or more banks or security brokerage firms that assemble a large block of the shares of a foreign corporation without the participation of the underlying corporation. Most DRs issued by Taiwanese listing companies are sponsored DRs.

The first issuance of DRs sponsored by a Taiwan company was the GDR of China Steel Corporation in May 1992. At the end of 1999, 36 Taiwanese listing companies have issued DRs and the total amount of issuance had reached to 6.243 billion US dollars. Among these DRs, high technology companies are the major sponsors, which record 22 or 61.11% of issuances. The capital raised by these high technology companies was 4.475 billion US dollars, or 71.68% of all issuances. Besides, the frequency of DRs issued by Taiwan companies dramatically increased from 1994 to 1999. The status of Taiwan-listed companies issuing DRs is summarized in Table 1.

In recent years, financial deregulation and international financial integration have resulted in a large amount of research on the dynamics of international transmission between GDRs or ADRs and their underlying securities. For example, Barclay et al. (1990) report that dual listing of sixteen New York Stock Exchange (NYSE) listed companies on the Tokyo Stock Exchange (TSE) has no impact on the variances of NYSE close-to-close returns on the stocks. Kato et al. (1991) and Wahab et al. (1992) try to find arbitrage opportunities between the prices of ADRs and underlying securities. They generally support the notation that, after transactions costs, few profitable opportunities exist in these markets, implying that both markets are efficient. Jayaraman, Shastri and Tandon (1993) suggest that the listing of ADRs are associated with permanent increases in the return volatilities of the underlying stocks. Kim, Szakmary and Mathur (2000) use both a vector autoregressive (VAR) model with a cointegration constraint and a seemingly unrelated regression (SUR) approach to examine the relative importance of, and the speed of adjustment of ADR prices to, these underlying factors. Their results show that the ADRs appear to initially overreact to the US market index but underreact to changes in underlying share prices and exchange rate.

Multiple listing offers a unique opportunity to study the transmission of pricing infor-mation across markets. Neumark, Tinsley and Tonsini (1991) find that the foreign market reacts to domestic price changes more quickly than the domestic market reacts to foreign

T able 1. Status of T aiw an listed compan y issuing sponsored DRs Common Number of Stock Listing Issuing DRs (in Unit Price T otal Amount Leading Corporation Name Code Industry Location Date thousand) (US dollar) (in thousand) Underwriter China Steel 2002 TT Steel Global 05/28/92 18,000 18.20 327,600 Goldman Sachs Asia Cement (1) 1102 TT Cement Global 06/23/92 2,400 27.50 66,003 Mor gan Stanle y President Enterprises 1216 TT F ood LX 11/24/92 4,993 16.51 82,426 CS First Boston Chia Hsin Cement 1103 TT Cement LI 05/25/93 2,100 16.90 35,490 Jarding Fleming T unte x Distinct 1462 TT T extile Global 05/04/94 7,000 12.12 84,840 Baring Brothers Microelectronics 2314 TT Electronic Global/LX 05/24/94 3,900 12.70 48,260 CS First Boston T echnology Hocheng 1810 TT Glass and LX 06/29/94 2,800 31.50 85,400 BZW Ceramic T ung Ho Steel 2006 TT Steel LX 08/09/94 6,000 17.20 103,200 Jarding Fleming Y ageo 2327 TT Electronic Global/LX 09/28/94 4,000 22.90 114,500 Schroder Aurora 2373 TT Electronic LX 01/27/95 1,875 16.00 30,000 N A GVC 2322 TT Electronic LI 04/03/95 5,000 15.30 76,500 Goldman Sachs ASE 2311 TT Electronic Global/LX 07/13/95 8,600 15.25 131,150 Mor gan Stanle y A.D.I. 2304 TT Electronic LX 09/28/95 2,500 16.96 42,400 Bank ers T rust W alsin Lihw a 1605 TT Elec. Cable Global/LX 10/03/95 10,000 12.18 121,800 Daiw a, Bank ers and W ire T rust Siliconw are Precision 2325 TT Electronic Global/LI 10/04/95 6,000 15.20 91,200 BZW Acer (1) 2306 TT Electronic Global/LI 11/01/95 17,000 12.99 220,830 Nomura Int ’l Macronix Int ’l 2337 TT Electronic N ASD A Q 05/14/96 10,000 17.76 176,700 CS First Boston Ev er green Marine 2603TT Marine Global/LI 07/30/96 10,800 18.05 194,940 Goldman Sachs Asia Cement (2) 1102 TT Cement Global/LI 09/12/96 3,750 20.00 60,000 SBC W arb ur g (continued )

T able 1. (Continued ) Common Number of Stock Listing Issuing DRs (in Unit Price T otal Amount Leading Corporation Name Code Industry Location Date thousand) (US dollar) (in thousand) Underwriter Lite-on T echnology 2346 TT Electronic LI 09/25/96 4,900 14.55 71,295 BZW Y ung Ming Marine 2609 TT Marine LI 11/14/96 10,000 11.64 116,392 UBS Accton T echnology 2345 TT Electronic Europe/LI 02/01/97 12,000 7.51 90,120 Jarding Fleming T eco Elec. & Mach. 1504 TT Engineer Global/LI 03/27/97 5,540 20.08 111,241 SBC W arb ur g Asustek Computer 2357 TT Electronic Global/LI 05/30/97 21,000 11.23 235,830 Nomura Int ’l Standard F oods 1227 TT F ood Global/LX 06/19/97 3,000 9.69 29,070 Schroder Synne x T echnology (1) 2347 TT Electronic Global/LX 07/03/97 431 9.81 4,225 Baring Brothers Synne x T echnology (1) 2347 TT Electronic Global/LX 07/03/97 6,270 22.23 139,382 Baring Brothers Acer (2) 2306 TT Electronic Global/LI 07/23/97 10,000 16.06 160,600 Goldman Sachs TSMC (1) 2330 TT Electronic NYSE 10/08/97 24,000 24.78 594,720 Goldman Sachs Fubon Insurance 2817 TT Insurance Global/LI 04/17/98 8,000 20.07 160,560 CS First Boston D-Link 2332 TT Electronic Global/LX 09/18/98 5,000 10.13 50,650 Salomon Smith Barne y W inbond (1) 2344 TT Electronic LX 02/05/99 14,600 11.45 167,170 ABN Amro Acer Peripherals 2352 TT Electronic LI 06/29/99 2,700 23.22 62,694 Nomura Int ’l TSMC (2) 2330 TT Electronic NYSE 07/15/99 12,094 24.00 296,500 Goldman Sachs Synne x T echnology (2) 2347 TT Electronic Global/LX 08/12/99 5,463 18.93 103,415 Jarding Fleming TSMC (3) 2330 TT Electronic NYSE 09/09/99 5,486 28.96 158,897 Goldman Sachs Mosel V itelic 2342 TT Electronic LX 09/16/99 9,980 8.70 86,867 Nomura Int ’l Hou Hai Precision 2317 TT Electronic LI 10/07/99 30,000 13.89 416,700 W arb ur g Dillion Read Ritek 2349 TT Electronic LX 10/15/99 27,500 11.86 326,150 ABN Amro Po werchip Semicond 5346 TT Electronic LX 10/21/99 27,000 10.70 288,900 Nomura Int ’l F ar East T extile 1402 TT T extile LX 10/25/99 13,500 14.00 189,000 Goldman Sachs Campal 2324 TT Electronic LX 11/09/99 8,000 15.27 122,160 Nomura Int ’l W inbond (2) 2344 TT Electronic LX 11/12/99 10,000 16.70 167,000 W arb ur g Dillion Read T otal 6,242,778 Data sour ce: Securities and Futures Commission Ministry of Finance, R.O.C.; LX: Lux embour g; LI: London.

price changes. This asymmetry, confirmed for price indices by Eun and Shim (1989) and Hamao, Masulis and Ng (1990), is interpreted by Garbade and Silber (1979) as evidence that the foreign market acts as a satellite to the domestic market. Hauser, Tanchuma and Yaari (1998) investigate five companies based in Israel whose stocks are listed on both the Tel Aviv Stock Exchange and NASDAQ. Their empirical tests of causality in price changes use the side-by-side Box-Jenkins ARIMA models and the Sims VAR model. Overall, the results show that price causality in dually listed stocks is unidirection from the domestic market to the foreign market. Jithendranathan, Nirmalanandan and Tandon (2000) evaluate market segmentation and its effect on the pricing of cross-listed securities using Indian Global De-positary Receipts (GDRs). They report that capital flow barriers existing in India lead to the GDRs being priced at a premium over the exchange rate adjusted prices of the underlying Indian securities. And GDR index returns are affected by both domestic and international factors, while the underlying Indian securities are affected only by domestic variables.

Earlier studies on international capital asset theory assume that international dually listed securities should sell at the same price in the absence of transaction costs and restriction to capital flows. Garbade and Silber (1979) reported that prices may differ between market centers for short intervals of time in imperfectly integrated market. The adjustment be-tween prices in market A and market B can be characterized in one of two ways: (1) the adjustment may be symmetrical; (2) the adjustment may be one-sided. Hence, this paper uses Granger tests to examine causal relations between the returns on GDRs or ADRs and their respective underlying Taiwanese securities. We use error-correction model to analyze the long run causal relations where the stock returns data is nonstationary. In addition, this paper further discusses the impact of premium or discount in overseas-listed stocks on the price transmission effect.1_{The net buy by QFIIs}2_{in Taiwan is also one of important factors} to be measured in price transmission effect because QFIIs play an increasingly important role in Taiwan market since the MSCI indices including this emerging market. So we also examine the effect of the net buy variable on causal relations.

Similar to the most existing research, our empirical results show the return causality is mostly unidirection from the domestic market to foreign market for dually listed Taiwan stocks. Besides, the prices of both markets will make adjustment to establish a long run cointegrated equilibrium. Unlike prior studies, this paper finds that both the premium or discount and QFII’s net buy have significant impacts on international price transmission for over twenty percent samples. Empirical tests show that our model is robust.

The remainder of this article is organized as follows: the next section presents the data description and the related empirical methodology. Our empirical results are described in section three. The final section concludes the paper.

2. Data and methodology 2.1. Data description

Thirty-six listed companies issued GDRs or ADRs by the end of 1999 in Taiwan; ninety-five GDRs or ADRs were issued and listed on exchange or over-the-counter (see Table 2). This

T able 2. List of T aiw an listed companies issuing GDRs or ADRs Bloomber g Code Corporation Code High Liquid Lo w Liquid No Quotation Issuing Date Shares of QFIIs ’ Holding (in thousand) Ratio of QFIIs ’ Holding (%) Ratio of QFIIs ’ & Non-QFIIs ’ Holding (%) 2002 TT CSGDS LI CHCG LX CISEY US 05/28/92 547,186 6.39 6.82 CNS GR (EUR$) CIS XF US 1102 TT A CGDS LI AIA GR (EUR$) 06/23/92 69,589 3.72 11.88 1216 TT PENT LX UPEZY US 11/24/92 190,148 6.50 13.62 1103 TT CHCDS LI CHSNF US 05/25/93 7,436 1.06 11.43 1462 TT TTXS LI TUNG LX 05/04/94 23,887 1.07 12.01 2314 TT METG LX 05/24/94 56,630 17.15 46.38 1810 TT HCDR LI HCDG LX 06/29/94 8,589 2.05 2.57 2006 TT THSGG LX 08/09/94 1,082 2.28 2.94 2327 TT Y A GG LX Y A G SP; YGEQY US 09/28/94 79,407 6.67 16.43 2373 TT A URD LI A URG LX A U R S P 01/27/95 38,810 6.24 9.43 2322 TT GVCD LI; 2552Q US 04/03/95 17,030 2.27 11.78 2311 TT ASED LI ASEG LX ASE SP 07/13/95 256,888 12.97 39.28 2304 TT ADOD LI ADIC LX 09/28/95 22,839 3.38 15.37 1605 TT WL WD LI WL WD LX 10/03/95 25,812 0.82 13.53 2325 TT SILD LI SLCZF US; SPIAF US; 10/04/95 110,158 9.77 16.34 SLCWY US 2306 TT A CID LI A CEHF US A CER Y U S 11/01/95 320,816 10.32 25.69 A CIG GR (EUR$) 1280Q US 2337 TT MXICY US MSID LI MXITY US 05/14/96 201,890 9.49 19.91 MXIC GR (EUR$) MXIA LI 2603 TT EGMD LI 07/30/96 119,344 6.43 33.14 EMA GR (EUR$) 2346 TT L TTD LI 09/25/96 32,403 6.57 7.04 2609 TT YMTD LI 1848Q US 11/14/96 175,538 10.45 11.52 2345 TT A T OD LI A CTG LX; A CTVF US; 02/01/97 26,491 11.29 13.69 A THYYP US

1504 TT TECD LI 03/27/97 35,806 2.05 4.99 2357 TT ASKD LI 05/30/97 200,007 17.45 17.73 1227 TT SFTD LI 06/19/97 21,007 6.95 12.31 2347 TT SYXD LI SYXTY LX; SYXTY US; 07/03/97 36,112 10.32 37.87 2678Q US 2330 TT TSM US 10/08/97 1,777,349 17.79 34.69 TMSD LI TSF A G R (EUR$) 2817 TT FBND LI FUIZF US; FUISY US; 04/17/98 86,170 5.00 9.57 2030Q US 2332 TT DLKD LI DLNKG LX 09/18/98 90,727 29.63 31.99 2344 TT WBDD LI WBD A LX; WBEKY US 02/05/99 272,733 7.73 12.20 2352 TT A CER LX; A CED LI 06/29/99 137,787 15.41 15.81 2342 TT MSVD LI 2671Q US 09/16/99 57,125 2.25 14.26 2317 TT HHPD LI HNHPF US; 1092Z LX; 10/07/99 326,998 29.73 34.53 2685Q US 2349 TT RKCD LI R T K G R (EUR$) RITK LX 10/15/99 12,776 2.01 3.42 5346 TT POSD LI POSD LX 10/21/99 303,432 17.84 39.16 1402 TT FETD LI F ARE LX; 2699Q US 10/25/99 460,977 16.76 23.76 2324 TT CPED LI CPED LX 11/09/99 152,969 9.83 10.06 A v erage 175,110 9.10 17.87 Data sour ces: Bloomber g Information System; Securities and Futures Commission Ministry of Finance, R.O.C.; EnT rust Securities Compan y.

Table 3. List of selected samples

Bloomberg Code

Corporation GDRs or Issuing Sample

Code ADRs Code Date

Shares of QFIIs’ Holding (in thousand) Ratio of QFIIs’ Holding (%) Ratio of QFIIs’ & Non-QFIIs’ Holding (%) Days 2317 TT HHPD LI 10/07/99 326,998 29.73 34.53 151 5346 TT POSD LI 10/21/99 303,432 17.84 39.16 141 2330 TT TSM US 10/08/97 1,777,349 17.79 34.69 623 2357 TT ASKD LI 05/30/97 200,007 17.45 17.73 626 1402 TT FETD LI 10/25/99 460,977 16.76 23.76 139 2311 TT ASED LI 07/13/95 256,888 12.97 39.28 626 2345 TT ATOD LI 02/01/97 26,491 11.29 13.69 626 2609 TT YMTD LI 11/14/96 175,538 10.45 11.52 626 2306 TT ACID LI 11/01/95 320,816 10.32 25.69 626 2347 TT SYXD LI 07/03/97 36,112 10.32 37.87 626 2324 TT CPED LI 11/09/99 152,969 9.83 10.06 128 2325 TT SILD LI 10/04/95 110,158 9.77 16.34 626 2337 TT MXICY US 05/14/96 201,890 9.49 19.91 539 2344 TT WBDD LI 02/05/99 272,733 7.73 12.20 307 1227 TT SFTD LI 06/19/97 21,007 6.95 12.31 626 2346 TT LTTD LI 09/25/96 32,403 6.57 7.04 626 2603 TT EGMD LI 07/30/96 119,344 6.43 33.14 626 2002 TT CSGDS LI 05/28/92 547,186 6.39 6.82 626 1102 TT ACGDS LI 06/23/92 69,589 3.72 11.88 626 2342 TT MSVD LI 09/16/99 57,125 2.25 14.26 162 1605 TT WLWD LI 10/03/95 25,812 0.82 13.53 626 Average: 261,658 10.71 20.73 Sum: 10,328

paper chooses twenty-one GDRs or ADRs from the ninety-five, representing twenty-one listed companies, for total data of 10,328 sample days (see Table 3). The principal of making selections is as below:

1. Forty-two GDRs or ADRs which have no quotations and no trading are eliminated. Eighteen GDRs or ADRs which have quotations but traded light are also excluded. 2. We deleted some listed company samples because their shares held by QFII are less

than or equal to 6.3%. Exceptions are Asia Cement, Mosel Vitelic, and Walsin Lihwa, because these stocks are included in MSCI Taiwan Index.

3. We chose GDRs or ADRs on exchange for some companies issuing more than one and being liquid, for example, Taiwan Semiconductor Manufacturing (TSM US). GDRs or ADRs in dollar quotation are selected if they have dollar and eurdollar quotations, for example, ACID LI, EGMD LI, and CSGDS LI.

4. Finally D-Link is also deleted because of late issuance and there only being forty-seven collected days.

The data for this paper are taken from three sources: the daily close price for GDRs or ADRs and NT exchange rate are collected from the Bloomberg information system; the

underlying stock close price and adjusted price for ex-dividend are provided by Taiwan Economic Journal (TEJ); the volume of net buy by foreign institutions is offered by the EnTrust Securities Company.

The period of selected samples is from October 8, 1997 to May 31, 2000 on the basis of the three points below:

1. In September 1997, Taiwan stocks were first included in the MSCI indices as a result of increasing foreign institution investment in the Taiwan equity market. It happened that the Asian Financial crisis began at the same time, which sharply increased systematic risk in Taiwan stock market. This event didn’t end until the fourth quarter of 1997. 2. Taiwan Semiconductor Manufacturing (TSM) ADR made an epochal entry for Taiwan

companies issuing GDRs or ADRs because TSM is at the head of Taiwan’s high-tech industry, the first Taiwan company listed in NYSE, and the issuing amount being the largest among all GDRs or ADRs.

3. These two years and seven months of selected samples cover a bearish market in 1998 and bullish market in 1999 and 2000. So this period represents a complete business cycle in the Taiwan stock market.

2.2. Methodology

The methodology employed in this study is based on Granger (1969). Other causality testing methods reported in the literature include the test proposed by Sims (1972) and the procedure suggested by Pierce and Haugh (1977). However, Granger’s tests are employed because they are superior to Sims’ (see Geweke, Meese and Dext (1983), and according to Hardouvelis (1988)), they perform well for small samples. However, it is necessary to test if the variables are stationary or not before Granger tests. If they are nonstationary, it is appropriate to specify by means of the vector error-correction models (Engle and Granger, 1987) to explore Granger causality relationship between GDRs or ADRs and the prices of underlying shares.

2.2.1 Unit root test. The assumptions of the classical regression model necessitate that the time series be stationary and the errors have a zero mean and finite variance. In the presence of nonstationary variables, there might be what Granger and Newbold (1974) call a spurious regression.3_{Thus, the first step in the analysis is to check if the structure of the} returns series is stationary by the augmented Dickey-Fuller test (ADF).

The augmented Dickey-Fuller test can be applied both in the case of a lower and a higher autoregressive (AR) process. The following equation presents a higher AR process version (with a constant and a time trend) of the Dickey Fuller test:

yt = a0+ a1t+ γ yt₋₁+ P 

i=2

where ytrepresents a time series, implies first difference, and t is the time trend. According to Said and Dickey (1984) the ADF test procedure is valid for a general ARMA process in the errors. The null hypothesis in the ADF test is unit root(γ = 0). For yt to be stationary, γ should be negative and significantly different from zero.

2.2.2 Cointegration tests. A system of nonstationary individual stock price in levels can, however, share common stochastic trends. Put simply, two nonstationary time series are cointegrated if a linear combination of two variables is stationary, that is, converges to an equilibrium over time. The main idea behind cointegration is a specification of models that includes beliefs about the movements of variables relative to each other in the long-run, such as the price of Taiwan’s stock and GDRs or ADRs. Thus a common stochastic trend in a system of stock prices can be interpreted to mean that the stochastic trend in Taiwan’s stock price is related to the GDRs or ADRs trend. There exists more than one method of conducting cointegration tests. The long-run relationship tests in this paper are conducted by means of the method developed by Johansen (1988) and Johansen and Juselius (1990). The Johansen maximum likelihood approach sets up the nonstationary time series as a vector autoregressive (VAR). The model is also called vector error-correction model (VECM):

Xt = c + N  i₌₁ iXt−i+  Xt−1+ ηt, ηt ∼ niid(0, δ)

where Xt is a vector of nonstationary (in levels) variables, implies first difference and c is the constant term. The information on the coefficient matrix between the levels of the

seriesis decomposed as= αβ where the relevant elements of theα matrix are the

adjustment coefficients and theβ matrix contains the cointegrating vectors. α and β are p× r matrices of full rank. If r = 0, then= 0, and there exists no linear combination of the elements of Xt that is stationary. At the other extreme, if rank(



) = p, Xtis itself a stationary process. In the intermediate case, when 0< r < p, there exist r stationary linear combinations of the elements of Xt. The constant term is included to capture the trending characteristic of the time series involved. The Johansen method provides the trace test and to determine the number of cointegrating vector. It is defined as:

trace statistic= −T p  i_=r+1

ln(1 − λi)

for r = 0, 1, 2, . . . , p − 1 where λi is the i th largest eigenvalue. The critical values for the trace statistic are reported by Osterwald-Lenum (1992), not those tabulated in Johansen and Juselius (1990). The trace statistic generally has greater power when theλi sare evenly distributed.

2.2.3 Granger causality test. The objective of this section is to investigate causal relations between the returns on GDRs or ADRs and their respective underlying Taiwan securities. The methodology employed in this study is based on Granger (1969). The Granger Causality

tests with difference form involve the estimation of the following equation: Rif,t = λ f 0 + λ f 1R f i,t−1+ λ f 2R d i,t+ ε f i,t (1) Rid,t = λ d 0+ λ d 1R d i,t−1+ λ d 2R f i,t−1+ ε d i,t (2) where Ri,t= log(p f(d) i_,t ) − log(p f(d) i_,t ) pif,t(d): p f

i,t denotes the close price of GDRs or ADRs in day t; pdi,t denotes the close price of the underlying security in day t.

Within the same calendar day, Taiwan market closes earlier than US and European markets, so foreign investors can observe the returns of both markets on the same day. On the other hand, the domestic investors observe the returns of the preceding day overseas. The regression model is set up so that Rtfregresses on R

f t−1and R d t, but R d t regresses on R d t−1and Rtf₋₁. The Granger causality tests with the VECM involve the estimation of the following equation while the data exist cointegration relationship:

Rif_,t = k f 0 + k f 1R f i_,t−1+ k f 2R d i,t+ k f 3  log pif_,t−1− c0− c1log pdi,t−1  + εf i_,t (3) Rid,t = k d 0 + k d 1R d i,t−1+ k d 2R f i,t−1+ k d 3  log pid,t−1− d0− d1log p f i,t−1  + εd i,t (4)

The lambdas and kappas are the parameters to be estimated. In the above estimation of equation (1) to equation (4), if the estimated coefficientsλ₂f and k₂f of equations (1) and (3) are statistically significant while the estimated coefficientsλd

2and k

d

2 of equations (2) and (4) are not statistically significant, then the results suggest a uni-directional causality, in the Granger sense, from the Taiwan stock returns to change GDR stock returns. In terms coined by Garbade and Silber (1979), the underlying security market is dominant and the overseas security market is a satellite. If, on the other hand, the estimated coefficientsλd

2 and k2d of equations (2) and (4) are statistically significant while the estimated coefficientsλ₂f and k₂f of equations (1) and (3) are not statistically significant, then uni-directional causality exists from changes in GDRs or ADRs to Taiwan’s stock returns. If the four coefficients are statistically significant in equations (1) and (4), then the data provide evidence of bi-directional causality. Absence of bi-directional causality is indicated when the set of parameters λd 2, k d 2,λ f 2 and k f

2 are statistically insignificant. Finally, both k

d

3 and k

f

3 represent the speed of adjustment coefficient for reflecting the long-run disequilibrium in the prices between the underlying stock and the GDRs or ADRs. c0, d0, c1, d1are cointegrating coefficients.

2.2.4 The impact of premium and net buy on Granger causality

The premium effect. Several studies on the pricing behavior of dual listed international securities do not find any significant difference between the domestic price and exchange rate adjusted price of the same security listed in an overseas market.4 Park and Tavakkol

(1994) examine the exchange rate adjusted returns of Japanese ADRs and the underly-ing stocks and find no significant differences between the two. On the other hand, Miller and Morey (1996) find intra-day pricing differences between the British ADRs and the underlying securities. This paper examines whether the price transmission effect is signifi-cantly amplified while the magnitude of the premium or discounts in overseas-listed stocks increases. We set up the estimation of the following equations:

Rif,t = λ f 0 + λ f 1R f i,t−1+  βf 0 + β f 1 ∗ prem f t  Rdi,t+ ε f i,t (5) Rid_,t = λ d 0+ λ d 1R d i_,t−1+  βd 0 + β d 1 ∗ prem d t₋₁  Rif,t−1+ ε d i_,t (6) Rif_,t = k f 0 + k f 1R f i_,t−1+  φf 0 + φ f 1 · prem f t  Rdi_,t + kf 3  log pif_,t−1− c0− c1log pid_,t−1  + εf i_,t (7) Rid_,t = k d 0 + k d 1R d i_,t−1+  φd 0 + φ d 1 · prem d t  Rif_,t−1 + kd 3  log pd_i,t−1− d0− d1log p f i_,t−1  + εd i,t (8) where premtf = pif,t−1· EXt− pid,t pd i_,t (9) premd_t = p f i_,t−1· EXt−1− pid_,t−1 pd i_,t−1 (10) EXt: exchange rate in day t

We further discuss the above equations in two cases: one, both the premium and Rd

t have the same sign. For example, if foreign investors observe that there exists positive premium and the underlying stock price rises, then through the market mechanism of arbitrary transactions, the prices of ADRs or GDRs will not change or even go down in order to shrink the price gap between the underlying stock and ADRs or GDRs. Therefore, bothβ₁f andφ₁f will be expected to be equal to or less than zero. Second, the premium and Rd

t have the opposite sign. In this case, if foreign investors find that there exists positive premium but the underlying stock price is down, then the prices of ADRs or GDRs will decline. Thus, bothβ₁f andφ₁f will be expected to be greater than or equal to zero. According to the above, if both the premium and Rtf−1are the same sign, for example, domestic investors observe that there exists positive premium and ADRs or GDRs price is also up, then the underlying stock price will go up and bothβ₁dandφ₁dwill be expected to be greater than or equal to zero. On the contrary, when both the premium and Rtf−1are the opposite sign, for example, domestic investors observe that there exists positive premium but ADRs or GDRs price is down, the underlying stock price will not change or even go up in order to shrink the price gap. Therefore, bothβ₁dandφ₁d will be expected to be equal to or less than zero.

Net buy effect. During these years QFIIs are more aggressive to increase their investment positions in Taiwan. In consequence of the substantial growth of QFIIs’ portfolio holding in Taiwan’s equities, the net buy or net sell of QFIIs’ daily trading has become an important investment signal for all investors participating in the Taiwan market. Stock with positive net buy by QFIIs is often followed by an increase in share price on next trading day, in particular, for those stocks underlying ADRs or GDRs. Thus, we also add the variable of QFIIs’ daily net buy into our empirical models to further examine the impact of net buy on Granger causality relationship between the underlying stock and the ADRs or GDRs. The models are as follows:

Rif,t = λ f 0 + λ f 1R f i,t−1+  δf 0 + δ f 1 · BS f t  Rid,t+ ε f i,t (11) Rid,t = λ d 0+ λ d 1R d i,t−1+  δd 0 + δ d 1 · BS d t−1  Rif,t−1+ ε d i,t (12) Rif,t = k f 0 + k f 1R f i,t−1+  ωf 0 + ω f 1 · BS f t  Rid,t + kf 3  log pif,t−1− c0− c1log pid,t−1  + εf i,t (13) Rid,t = k d 0 + k d 1R d i,t−1+  ωd 0 + ω d 1· BS d t−1  Rif,t−1 + kd 3  log pdi_,t−1− d0− d1log pif,t−1  + εd i_,t (14) where

BStf: the volume of net buy for underlying stock by QFIIs in Taiwan observed by foreigners in day t; i.e., the net shares of total bought minus total sold for underlying stock by all QFIIs in day t

BSd

t₋₁: the volume of net buy for underlying stock by QFII in Taiwan observed by the domestic investors in day t

In the above estimation of equations (11) and (13), if QFIIs buy net shares for the underlying stock in Taiwan’s market and at the same time the return of the underlying stocks rises, the prices of GDRs or ADRs will rise, fall or hold steady, so the signs ofδ₁f and w₁f can’t be determined. There are two reasons to explain the above phenomenon. One is that the prices of the underlying stocks are relatively undervalued as a consequence of QFIIs’ net buy in order to get arbitrage profit. Of course the gap between both prices can be shrunk by arbitrage trading. The other reason is that the prospect of the underlying company demonstrates such potential that QFIIs purchase these underlying stocks, and as a result the prices of the underlying stocks and ADRs or GDRs simultaneously rise. On the other side, in equations (12) and (14), if the return of GDRs or ADRs is negative and QFIIs buy net shares of the underlying stock, the return of the underlying stock will be positive, negative or zero, so the signs of δ₁d and wd₁ are also uncertain. The reason for the uncertainty is that the prices of the underlying stocks are undervalued and QFIIs buying net shares will lead to the price of the underlying stocks rising. However, the underlying stock prices may also drop to reflect the falling price of GDRs or ADRs.

Robust test. Finally, our model simultaneously includes the premium and net buy factors to examine if the model is robust. This is reflected in the estimation of equations (15) to (18):

Rif,t = λ f 0 + λ f 1R f i,t−1+  θf 0 + θ f 1 · prem f t + θ f 2 · BS f t  Rid,t+ ε f i,t (15) Rid,t = λ d 0+ λ d 1R d i,t−1+  θd 0 + θ d 1 · prem d t + θ d 2 · BS d t−1  Rif,t−1+ ε d i,t (16) Rif,t = k f 0 + k f 1R f i,t−1+  γf 0 + γ f 1 · prem f t + γ f 2 · BS f t  Rdi,t + kf 3  log pif,t−1− c0− c1log pid,t−1  + εf i,t (17) Rid,t = k d 0 + k d 1R d i,t−1+  γd 0 + γ d 1 · prem d t + γ d 2 · BS d t−1  Rif,t−1 + kd 3  log pdi_,t−1− d0− d1log pif,t−1  + εd i_,t (18) 3. Empirical results

3.1. Unit root and cointegration tests results

As stated earlier, all series are applied in logarithmic form. As required of all cointegration tests, the series of stock price must first be inspected for the presence of unit roots. Table 4

Table 4. Augmented Dickey-Fuller tests for a unit roota

Corporation Code Taiwan GDRs or ADRs

1102 −2.721650 −2.885068 1227 −2.532691 −2.382961 1402 −1.494700 −1.252938 1605 −1.586543 −1.845639 2002 −2.816041 −2.404293 2306 −1.864972 −2.064231 2311 −1.910316 −2.050046 2317 −1.981890 −1.352853 2324 −1.791862 −1.935491 2325 −3.505806** −3.729858** 2330 −1.978247 −2.587620 2337 −1.401737 −1.530777 2342 −2.154823 −2.252726 2344 −3.093396 −3.008204 2345 −2.522986 −2.494403 2346 −1.172046 −1.113604 2347 −3.763689** −3.686703** 2357 −2.081670 −2.191681 2603 −2.337686 −1.714761 2609 −3.341800** −3.717536** 5346 −2.465527 −2.501067

a_{The entry in each cell is the ADF statistic.}

Table 5. Cointegration test results (Lags in the VAR= 1)

ra_{= 0 r=}< 1

Corporation Code Eigenvalue Trace Statistic Eigenvalue Trace Statistic

1102 0.028652 26.32853*** 0.013037 8.188700*** 1227 0.018837 12.46983 0.000966 0.603329 1402 0.071200 12.34016 0.016081 2.221029 1605 0.034779 23.29036*** 0.001924 1.202012 2002 0.031511 24.74242*** 0.007604 4.763040** 2306 0.199605 14.02069*** 0.002038 1.273044 2311 0.033467 24.49023*** 0.005194 3.249319 2317 0.090624 16.04527** 0.020229 2.840711 2324 0.083378 14.61861 0.028546 3.649078 2330 0.044065 28.63125*** 0.001040 0.645971 2337 0.215139 13.00895*** 0.000003 0.001861 2342 0.114581 20.31530*** 0.005262 0.844201 2344 0.065481 20.70226*** 0.000153 0.046571 2345 0.011082 8.226734 0.002038 1.272949 2346 0.013422 11.19552 0.004419 2.763638 2357 0.047425 32.22369*** 0.003050 1.906004 2603 0.022752 18.79700** 0.007083 4.435474** 5346 0.060651 10.65297 0.013973 1.955989

a_{r is hypothesized number of cointegrating relationships.}

*** and **imply rejection of the null at 1% and 5% level, respectively.

presents the results from ADF tests. We employed Akaike’s information criterion to select the appropriate lag lengths. For most of the series, we are unable to reject the unit root hypothesis, but there are some exceptions. In other words, most of time series data are I (1), but some series data are I (0). When the series are stable, they don’t need the cointegration test.

Table 5 presents the results from the cointegration tests. In this paper, we used trace test to determine the number of cointegrating vectors. Results show a cointegration relationship existing between most Taiwan stock prices and their GDR’s or ADR’s prices. In other words, we are able to find a stationary long run relationship between both. So, we should employ VECM to test Granger causality relationship for existing cointegration time series. It is appropriate to simultaneously consider long- and short-term effects. For time series without cointegration, they directly take the first difference form to test Granger causality relationship. As reported in Table 5, trace tests indicate that at least one cointegration relationship exists for twelve of the firms. Each firm’s cointegrating vector is calculated and incorporated in the VAR model estimation to capture the long run equilibrium relationship.

3.2. Granger causality tests results

Table 6 shows the results from Granger Causality with difference form. Only seven samples have bi-directional causality but fourteen samples among twenty-one total show

Table 6. Granger causality tests results—difference form R_i,tf = λ₀f+ λ₁fR_i,t−1f + λ₂fRd i,t+ εi,tf Rd i,t= λd0+ λ d 1R d i,t−1+ λd2R f i,t−1+ εi,td λf 0 λ f 1 λ f 2 Corporation Code D. V.a _λd 0 λ d 2 λ d 1 1102 Rtf −0.0002 −0.0381 0.8993 (0.7954) (0.1667) (0.0000)*** Rd t −0.0005 0.1077 −0.0446 (0.6563) (0.0241)** (0.4457) 1227 Rtf −0.0002 0.0844 0.7185 (0.8390) (0.0073)*** (0.0000)*** Rd t −0.0006 0.1447 0.0154 (0.5680) (0.0010)*** (0.7632) 1402 Rtf 0.0001 0.0508 0.7535 (0.9609) (0.4282) (0.0000)*** Rd t 0.0008 0.1383 −0.0683 (0.8085) (0.1808) (0.5582) 1605 Rtf 0.0001 −0.0282 0.9315 (0.9272) (0.1644) (0.0000)*** Rd t 0.0005 0.0931 −0.1169 (0.6575) (0.2052) (0.1407) 2002 Rtf −0.0000 0.0164 1.0482 (0.9792) (0.5628) (0.0000)*** Rd t 0.0002 0.1391 0.1770 (0.8430) (0.0003)*** (0.0018)*** 2306 Rtf 0.0000 −0.1074 0.9934 (0.9536) (0.0000)*** (0.0000)*** Rd t 0.0009 0.1744 −0.1645 (0.5259) (0.0146) (0.0422)** 2311 Rtf 0.0001 0.0156 0.9501 (0.9152) (0.5719) (0.0000)*** Rd t 0.0004 0.1021 −0.1039 (0.7498) (0.0216) (0.0742)∗ 2317 Rtf −0.0003 0.1373 0.8626 (0.8942) (0.0305)** (0.0000)*** Rd t 0.0021 0.1098 0.0003 (0.4069) (0.1669) (0.9975) 2324 Rtf 0.0002 0.0530 1.0029 (0.9188) (0.2725) (0.0000)*** Rd t −0.0006 0.0938 −0.1403 (0.8474) (0.5136) (0.4079) 2325 Rtf 0.0000 −0.0440 0.9086 (0.9892) (0.1629) (0.0000)*** Rd t 0.0003 0.0594 −0.0370 (0.8375) (0.0896)* (0.4684) 2330 Rtf 0.0009 −0.1244 0.7192 (0.5526) (0.0008)*** (0.0000)*** Rd t 0.0011 0.2096 −0.0637 (0.3371) (0.0000)*** (0.1398)

Table 6. (Continued) λf 0 λ f 1 λ f 2 Corporation Code D. V.a _λd 0 λ d 2 λ d 1 2337 Rtf 0.0005 −0.1373 0.7759 (0.7222) (0.0000)*** (0.000)*** Rd t 0.0011 0.1300 −0.0548 (0.5018) (0.0080)** (0.3338) 2342 Rtf 0.0007 −0.0334 0.9720 (0.6420) (0.3512) (0.0000)*** Rd t 0.0056 0.0230 0.0473 (0.0815)* (0.8873) (0.7873) 2344 Rtf 0.0004 −0.0840 1.0020 (0.7229) (0.0074) (0.0000)*** Rd t 0.0033 0.1238 −0.1679 (0.0767) (0.1577) (0.1094) 2345 Rtf −0.0006 −0.0753 1.0214 (0.4457) (0.0001)*** (0.0000)*** Rd t 0.0004 0.0116 0.0874 (0.7739) (0.8720) (0.2932) 2346 Rtf −0.0004 −0.0876 0.9839 (0.6363) (0.0000)*** (0.0000)*** Rd t −0.0000 0.0091 0.1163 (0.9796) (0.9094) (0.1848) 2347 Rtf 0.0000 −0.1148 0.9800 (0.9753) (0.0000)*** (0.0000)*** Rd t 0.0004 0.0769 0.0748 (0.7415) (0.2383) (0.3125) 2357 Rtf 0.0003 −0.0628 1.0568 (0.7993) (0.0355)** (0.0000)*** Rd t 0.0010 0.1374 −0.0739 (0.3419) (0.0001)*** (0.1711) 2603 Rtf −0.0000 0.0439 0.8715 (0.9898) (0.1034) (0.0000)*** Rd t −0.0002 0.0722 0.0435 (0.8058) (0.1542) (0.4660) 2609 Rtf 0.0006 −0.0094 0.8723 (0.4126) (0.7111) (0.0000)*** Rd t −0.0001 0.0642 −0.0199 (0.9626) (0.2578) (0.7549) 5346 Rtf 0.0021 −0.0297 0.9594 (0.2426) (0.4854) (0.0000)*** Rd t 0.0033 0.0259 0.0199 (0.3150) (0.8617) (0.9020)

a_{D. V. represents dependent variable.}

*Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.

uni-directional causality. This demonstrates that most of the time the Taiwan stock re-turns change GDRs or ADRs rere-turns. In addition, the estimated coefficientλd₂ approaches 100 percent, despite the estimated coefficientλ₂f being near to zero. So our empirical re-sults mostly support Garbade and Silber’s (1979) idea that the underlying security market is dominant and the overseas security market is a satellite.

To continue, the cointegration relationship data employ the Granger causality test with VECM which simultaneously considers long run and short run effects between the returns of Taiwan stock and GDRs or ADRs. Table 7 presents the empirical results with VECM. In the short run, it shows similar results that Taiwan stock returns significantly influence GDRs or ADRs returns but not the reverse. In the long run, all the coefficients for the speed of Table 7. Granger causality tests results—VECM

R_if_,t = k₀f + k₁fR_if_,t−1+ k₂fRd_i,t+ k₃f(log p_if_,t−1− c0− c1log pdi,t−1) + εif,t Rd i,t = kd0+ k1dRid,t−1+ kd2R f i,t−1+ kd3(log pid,t−1− d0− d1log pif,t−1) + εdi,t κf 0 κ f 1 κ f 2 κ f 3 Corporation Code D.Va _κd 0 κ d 2 κ d 1 κ d 3 1102 Rtf −0.0002 −0.0240 0.9088 −0.0545 (0.8056) (0.3817) (0.0000)*** (0.0000)*** Rd t 0.0927 −0.0324 −0.0302 (0.0558)* (0.5812) (0.0857)* 1605 Rtf 0.0038 −0.0274 0.9311 −0.0007 (0.6412) (0.1784) (0.0000)*** (0.6455) Rd t 0.0187 0.0925 −0.1125 −0.0064 (0.2620) (0.2078) (0.1565) (0.2749) 2202 R_tf −0.0000 0.0277 1.0570 −0.0551 (0.9763) (0.3236) (0.0000)*** (0.0000)*** Rd t 0.0002 0.1333 −0.1699 −0.0146 (0.8434) (0.0007)*** (0.0031)*** (0.3940) 2306 Rtf −0.0001 −0.0062 1.0453 −0.5855 (0.8883) (0.7336) (0.0000)*** (0.0000)*** Rd t 0.0009 −0.0207 0.0012 −0.3905 (0.5105) (0.8064) (0.9889) (0.0000)*** 2311 Rtf −0.0001 0.0318 0.9659 −0.0425 (0.9250) (0.2527) (0.0000)*** (0.0004)*** Rdt 0.0004 0.0803 −0.0930 −0.0437 (0.7441) (0.0724)* (0.1084) (0.0008)*** 2317 Rtf −0.0005 0.1613 0.8939 −0.0467 (0.8620) (0.0134)** (0.0000)*** (0.1231) Rdt 0.0021 0.0461 0.0350 −0.1167 (0.3888) (0.5669) (0.7434) (0.0047)*** 2330 Rtf 0.0009 −0.1134 0.7351 −0.0191 (0.5712) (0.0028)*** (0.0000)*** (0.1663) Rd t 0.0012 0.1740 −0.0648 −0.0607 (0.2966) (0.0000)*** (0.1255) (0.0000)***

Table 7. (Continued) κf 0 κ f 1 κ f 2 κ f 3 Corporation Code D.Va _κd 0 κ2d κd1 κ3d 2337 Rtf 0.0003 −0.0133 0.8312 −0.5796 (0.8124) (0.6487) (0.0000)*** (0.0000)*** Rdt −0.0048 0.0454 0.0288 −0.1681 (0.0790)* (0.4330) (0.6542) (0.0073)*** 2342 Rtf 0.0004 −0.0023 0.9973 −0.2655 (0.7952) (0.9475) (0.0000)*** (0.0000)*** Rdt 0.0057 −0.1091 0.1679 −0.2772 (0.0712)* (0.5285) (0.3598) (0.0418)** 2344 Rtf 0.0003 −0.0490 1.0169 −0.1630 (0.8205) (0.1131) (0.0000)*** (0.0000)*** Rd t 0.0006 0.0931 −0.1434 −0.0629 (0.8471) (0.3081) (0.1794) (0.2395) 2357 R_tf 0.0003 −0.0396 1.0849 −0.0663 (0.8356) (0.1880) (0.0000)*** (0.0001)*** Rd t 0.0010 0.1030 −0.0470 −0.0799 (0.3304) (0.0045)*** (0.3828) (0.0001)*** 2603 Rtf −0.0000 0.0362 0.8796 −0.0200 (0.9899) (0.1782) (0.0000)*** (0.0018)*** Rdt −0.0003 0.0691 0.0580 −0.0329 (0.8078) (0.1720) (0.3332) (0.0338)

a_{D. V. represents dependent variable.}

*Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level.

(.) represents p-value.

adjustment, k₃d’s and k₃f’s, are negative and most of them are significant. This result indicates that once the return relationship of the underlying stock and the GDRs or ADRs deviates from the long run cointegrated equilibrium, both markets will make opposite adjustment to reestablish the equilibrium in next period. In a word, Table 7 shows that price causality in dually listed stocks is mostly unidirectional from the domestic market to the foreign market and the prices of both markets will adjust to a long run cointegrated equilibrium.

3.3. Results of the premium and net buy effect

Table 8 reveals the results from Granger Causality involving the premium in overseas-listed stocks. The estimated coefficientβ₀f orφ₀f is significant and approaches to 1, but the estimated coefficientβ₀d orφ₀d is mostly not significant and near to zero. So the price transmission effect is still unidirectional from the domestic market to the foreign market. The transmission effect of seven samples is influenced by the premium in overseas-listed

T able 8. Change in Granger causality results — Premium R f i,t = λ f +0 λ f R1 f i,t− 1 + f β0 + β f ∗1 prem f t  dRi, t + ε f i,t (5 ) R d i,t = λ d+0 λ dR1 d i,t− 1 +  dβ0 + β d∗1 prem d t− 1 f Ri, t− 1 + ε d i,t (6 ) R f i,t = k f +₀ k f R1 f i,t− 1 + f φ0 + φ f ·prem1 f t  dRi, t + k f 3  log p f i,t− 1 − c0 − c1 log p d i,t− 1  + ε f _i,t (7 ) R d i,t = k d+0 k dR1 d i,t− 1 +  dφ0 + φ d·prem1 d t f Ri, t− 1 + k d 3  log p d i,t− 1 − d0 − d1 log p f i,t− 1  + ε d i,t (8 ) Corporation λ f (κ0 f )λ₀ f (κ1 f )β₁ f (φ₀ f )β0 f (φ₁ f )κ1 f ₃ Code D.V . a λ d(κ0 d)β0 d(φ0 d)λ0 d(κ1 d)β1 d(φ1 d)κ1 d 3 1102 R f t − 0. 0009 − 0. 0133 0.9580 − 1. 2071 − 0. 0572 (0.2970) (0.6279) (0.0000)*** (0.0003)*** (0.0000)*** R d t − 0. 0007 0.0347 − 0. 0107 0.6373 − 0. 0292 (0.5149) (0.5408) (0.8581) (0.0525)* (0.0961)* 1227 R f t − 0. 0007 0.0811 0.8006 − 0. 5703 (0.5076) (0.0098)*** (0.0000)*** (0.0187)** R d t − 0. 0006 0.1650 0.0157 − 0. 1647 (0.5884) (0.0017)*** (0.7584) (0.4745) 1402 R f t − 0. 0000 0.0522 0.7713 − 0. 1063 (0.9902) (0.4231) (0.0000)*** (0.8913) R d t 0.0003 − 0. 2416 − 0. 0094 1.5740 (0.9375) (0.1719) (0.9352) (0.0095)*** 1605 R f t 0.0038 − 0. 0274 0.9315 − 0. 0000 − 0. 0007 (0.6419) (0.1798) (0.0000)*** (0.9608) (0.6462) R d t 0.0188 0.1359 − 0. 1383 − 0. 7969 − 0. 0064 (0.2596) (0.1022) (0.0946)* (0.2653) (0.2750) 2002 R f t − 0. 0000 0.0294 1.0946 − 0. 0000 − 0. 0551 (0.9902) (0.2967) (0.0000)*** (0.4570) (0.0000)*** R d t 0.0001 0.0863 − 0. 1410 0.3727 − 0. 0105 (0.9458) (0.0770)* (0.0189)** (0.1098) (0.5458)

2306 R f t − 0. 0001 − 0. 0064 1.0349 0.0000 − 0. 5845 (0.8656) (0.7223) (0.0000)*** (0.5064) (0.0000)*** R d t 0.0001 − 0. 0315 1.0241 2.9732 − 0. 3836 (0.9620) (0.7080) (0.7873) (0.0118)** (0.0000)*** 2311 R f t 0.0004 − 0. 0927 0.0783 0.0101 − 0. 0437 (0.7520) (0.1106) (0.1973) (0.9601) (0.0029)*** R d t 0.0004 0.0783 − 0. 0927 0.0101 − 0. 0437 (0.7520) (0.1973)* (0.1106) (0.9601) (0.0028)*** 2317 R f t − 0. 0005 0.1583 0.8665 0.0001 − 0. 0460 (0.8467) (0.0162)** (0.0000)*** (0.6891) (0.1305) R d t 0.0023 0.0889 0.0365 − 0. 1182 − 0. 1148 (0.3696) (0.6091) (0.7340) (0.7808) (0.0062)*** 2324 R f t 0.0002 0.0665 0.9532 0.0004 (0.9002) (0.1722) (0.0000)*** (0.0948)* R d t − 0. 0005 0.1294 − 0. 1587 − 0. 7208 (0.8618) (0.4380) (0.3661) (0.6715) 2325 R f t 0.0000 − 0. 0375 0.9880 − 0. 0004 (0.9947) (0.2338) (0.0000)*** (0.0080)*** R d t 0.0004 0.0814 − 0. 0418 − 0. 0969 (0.7754) (0.1392) (0.4199) (0.6033) 2330 R f t 0.0007 − 0. 1210 0.6540 0.0002 − 0. 0192 (0.6663) (0.0016)*** (0.0000)*** (0.1832) (0.1633) R d t 0.0012 0.1850 − 0. 0653 − 0. 0288 − 0. 0607 (0.2822) (0.0002)*** (0.1230) (0.7821) (0.0000)*** 2337 R f t 0.0001 − 0. 0159 0.7948 0.0002 − 0. 5748 (0.9389) (0.5870) (0.0000)*** (0.0697)* (0.0000)*** R d t − 0. 0055 0.0428 0.0331 1.3587 − 0. 1723 (0.0450)** (0.4587) (0.6071) (0.0902)* (0.0059)*** (continued )

T able 8. (Continued ) Corporation λ f (κ0 f )λ₀ f (κ1 f )β₁ f (φ₀ f )β0 f (φ₁ f )κ1 f ₃ Code D.V . a λ d(κ0 d)β0 d(φ0 d)λ0 d(κ1 d)β1 d(φ1 d)κ1 d 3 2342 R f t 0.0004 0.0011 1.0185 − 0. 0002 − 0. 2645 (0.7845) (0.9739) (0.0000)*** (0.2407) (0.0000)*** R d t 0.0054 − 0. 0751 0.1661 1.5598 − 0. 2759 (0.0936)* (0.6732) (0.3657) (0.4087) (0.0431)** 2344 R f t 0.0003 − 0. 0491 1.0055 0.0000 − 0. 1631 (0.8121) (0.1126) (0.0000)*** (0.7325) (0.0000)*** R d t 0.0006 0.1234 − 0. 1593 − 0. 9562 − 0. 0694 (0.8447) (0.2073) (0.1418) (0.3852) (0.1991) 2345 R f t − 0. 0006 − 0. 0753 1.0178 0.0165 (0.4784) (0.0001)*** (0.0000)*** (0.8725) R d t 0.0005 0.0364 0.0835 − 0. 0992 (0.7587) (0.6630) (0.3166) (0.5583) 2346 R f t − 0. 0004 − 0. 0875 0.9915 − 0. 0002 (0.5862) (0.0000)*** (0.0000)*** (0.1940) R d t − 0. 0000 0.0073 0.1172 0.0202 (0.9794) (0.9331) (0.1893) (0.9564) 2347 R f t 0.0000 − 0. 1141 0.9831 − 0. 0001 (0.9570) (0.0000)*** (0.0000)*** (0.5909) R d t 0.0003 0.0632 0.0826 0.3805 (0.8264) (0.3601) (0.2725) (0.5483)

2357 R f t 0.0002 − 0. 0413 1.0726 0.0002 − 0. 0663 (0.8685) (0.1715) (0.0000)*** (0.4507) (0.0001)*** R d t 0.0011 0.1219 − 0. 0555 − 0. 1019 − 0. 0808 (0.2956) (0.0065)*** (0.3145) (0.4723) (0.0001)*** 2603 R f t − 0. 0000 0.0401 0.9174 − 0. 0002 − 0. 0210 (0.9859) (0.1361) (0.0000)*** (0.0522)* (0.0010)*** R d t − 0. 0003 0.0267 0.0798 0.2721 − 0. 0330 (0.7938) (0.6683) (0.2038) (0.2457) (0.0329)** 2609 R f t 0.0004 − 0. 0133 0.8217 0.0003 (0.5884) (0.5980) (0.0000)*** (0.0026)*** R d t − 0. 0001 − 0. 0067 − 0. 0011 0.3824 (0.9444) (0.9237) (0.9863) (0.0823)* 5346 R f t 0.0020 − 0. 0269 0.9496 0.0001 (0.2608) (0.5358) (0.0000)*** (0.7024) R d t 0.0036 0.0507 0.0435 1.1538 (0.2806) (0.7361) (0.7897) (0.2902) aD. V . represents dependent v ariable. *Signi fi cant at the 10% le v el. **Signi fi cant at the 5% le v el. ***Signi fi cant at the 1% le v el. (.) represents p -v alue.

T able 9. Change in Granger causality results — net b u y R f =_i,t λ f +0 λ f R1 f _i,t− 1 + f δ0 + δ f ·BS₁ f t  dRi, t + ε f i,t (11 ) R d i,t = λ d+0 λ dR1 d i,t− 1 +  dδ0 + δ d·BS1 d t− 1 f Ri, t− 1 + ε d i,t (12 ) R f =_i,t k f +₀ k f R₁ f _i,t− 1 +  ω f +0 ω f ·BS1 f t  dRi, t + k f ₃  log p f _i,t− 1 − c0 − c1 log p d i,t− 1  + ε f i,t (13 ) R d i,t = k d+0 k dR1 d i,t− 1 +  dω0 + ω d·BS1 d t− 1 f Ri, t− 1 + k d 3  log p d i,t− 1 − d0 − d1 log p f i,t− 1  + ε d i,t (14 ) Corporation λ f (κ0 f )λ₀ f (κ1 f )δ₁ f (ω0 f )δ₀ f (ω1 f )κ1 f ₃ Code D.V . a λ d(κ0 d)δ0 d(ω0 d)λ0 d(κ1 d)δ1 d(ω1 d)κ1 d 3 1102 R f t − 0. 0003 − 0. 0232 0.9114 0.0000 − 0. 0544 (0.7715) (0.3999) (0.0000)*** (0.6190) (0.0000)*** R d t − 0. 0007 0.0993 − 0. 0276 0.0000 − 0. 0321 (0.5349) (0.0403)** (0.6376) (0.0281)** (0.0671)* 1227 R f t − 0. 0002 0.0844 0.7185 − 0. 0000 (0.8445) (0.0075)*** (0.0000)*** (0.9931) R d t − 0. 0005 0.1395 0.0188 − 0. 0000 (0.6787) (0.0017)*** (0.7130) (0.3648) 1402 R f t 0.0010 0.0603 0.7471 − 0. 0000 (0.7254) (0.3541) (0.0000)*** (0.3768) R d t 0.0019 0.1136 − 0. 0581 − 0. 0000 (0.5792) (0.2830) (0.6192) (0.2884) 1605 R f t 0.0040 − 0. 0285 0.9311 − 0. 0001 − 0. 0007 (0.6288) (0.1618) (0.0000)*** (0.1673) (0.6381) R d t 0.0187 0.0934 − 0. 1131 0.0000 − 0. 0064 (0.2622) (0.2051) (0.1551) (0.8671) (0.2750)

2002 R f _t − 0. 0000 0.0287 1.0557 0.0002 − 0. 0545 (0.9375) (0.3058) (0.0000)*** (0.2859) (0.0000)*** R d t 0.0001 0.1260 − 0. 1600 − 0. 0001 − 0. 0126 (0.8751) (0.0014)*** (0.0056)*** (0.1506) (0.4643) 2306 R f _t − 0. 0002 − 0. 0062 1.0470 0.0001 − 0. 5870 (0.8095) (0.7317) (0.0000)*** (0.1820) (0.0000)*** R d t 1.0010 − 0. 0255 2.0044 − 0. 0001 − 0. 3961 (0.4729) (0.7628) (0.9609) (0.4134) (0.0000)*** 2311 R f _t 0.0003 0.0315 0.9717 − 0. 0001 − 0. 0429 (0.8138) (0.2575) (0.0000)*** (0.3295) (0.0004)*** R d t 0.0001 0.9502 0.0307 0.0001 − 0. 0425 (0.9524) (0.0000)*** (0.2709) (0.5249) (0.0004)*** 2317 R f _t 0.0005 0.1742 0.9476 − 0. 0005 − 0. 0506 (0.8558) (0.0068)*** (0.0000)*** (0.0104)** (0.0894)* R d t 0.0016 0.0297 0.0466 0.0002 − 0. 1176 (0.5108) (0.7112) (0.6613) (0.0746)* (0.0041)*** 2324 R f _t 0.0000 0.0649 1.0129 0.0004 (0.9906) (0.1700) (0.0000)*** (0.0059)*** R d t − 0. 0006 0.1539 − 0. 1924 − 0. 0002 (0.8449) (0.3160) (0.2739) (0.2659) 2325 R f _t 0.0004 − 0. 0462 0.9171 − 0. 0003 (0.8030) (0.1420) (0.0000)*** (0.0377)** R d t 0.0002 0.0590 − 0. 0365 0.0000 (0.8526) (0.0923)* (0.4746) (0.8057) (continued )

T able 9. (Continued ) Corporation λ f (κ0 f )λ₀ f (κ1 f )δ₁ f (ω₀ f )δ0 f (ω₁ f )κ1 f ₃ Code D.V . a λ d(κ0 d)δ0 d(ω0 d)λ0 d(κ1 d)δ1 d(ω1 d)κ1 d 3 2330 R f t 0.0010 − 0. 1158 0.7581 − 0. 0005 − 0. 0198 (0.5326) (0.0023)*** (0.0000)*** (0.0426)** (0.1499) R d t 0.0013 0.2031 − 0. 0812 − 0. 0004 − 0. 0595 (0.2348) (0.0000)*** (0.0521)* (0.0000)*** (0.0000)*** 2337 R f t 0.0002 − 0. 0127 0.8331 − 0. 0001 − 0. 5776 (0.8766) (0.6630) (0.0000)*** (0.4058) (0.0000)*** R d t − 0. 0048 0.0457 0.0286 0.0000 − 0. 1679 (0.0818)* (0.4318) (0.6582) (0.9477) (0.0075)*** 2342 R f t 0.0004 − 0. 0012 0.9929 0.0001 − 0. 2653 (0.7908) (0.9712) (0.0000)*** (0.4930) (0.0000)*** R d t 0.0056 − 0. 1101 0.1627 0.0001 − 0. 2724 (0.0758)* (0.5257) (0.3768) (0.5927) (0.0464)** 2344 R f t 0.0003 − 0. 0473 1.0162 0.0001 − 0. 1641 (0.8308) (0.1275) (0.0000)*** (0.4369) (0.0000)*** R d t 0.0008 0.0704 − 0. 1237 0.0002 − 0. 0598 (0.7932) (0.4528) (0.2538) (0.2943) (0.2638) 2345 R f t − 0. 0008 − 0. 0756 1.0191 0.0000 (0.3488) (0.0001)*** (0.0000)*** (0.2919) R d t − 0. 0001 − 0. 0117 0.1013 0.0001 (0.9473) (0.8719) (0.2227) (0.0196)**

2346 R f _t − 0. 0003 − 0. 0892 0.9828 0.0001 (0.6558) (0.0000)*** (0.0000)*** (0.4070) R d t − 0. 0000 0.0113 0.1124 0.0001 (0.9950) (0.8885) (0.2013) (0.5693) 2347 R f _t 0.0000 − 0. 1147 0.9797 0.0000 (0.9976) (0.0000)*** (0.0000)*** (0.8637) R d t 0.0004 0.0762 0.0751 0.0000 (0.7611) (0.2439) (0.3113) (0.8317) 2357 R f _t 0.0006 − 0. 0389 1.0834 − 0. 0002 − 0. 0643 (0.6252) (0.1950) (0.0000)*** (0.0858)* (0.0002)*** R d t 0.0009 0.1038 − 0. 0441 0.0001 − 0. 0770 (0.4311) (0.0042)*** (0.4128) (0.1710) (0.0003)*** 2603 R f _t − 0. 0000 0.0354 0.8802 0.0001 − 0. 0199 (0.9576) (0.1882) (0.0000)*** (0.5779) (0.0019)*** R d t − 0. 0002 0.0673 0.0594 − 0. 0000 − 0. 0325 (0.8204) (0.1866) (0.3233) (0.7620) (0.0363)** 2609 R f _t 0.0007 − 0. 0130 0.8748 − 0. 0002 (0.3496) (0.6081) (0.0000)*** (0.1131) R d t − 0. 0001 0.0670 − 0. 0226 0.0001 (0.9247) (0.2381) (0.7226) (0.2796) 5346 R f _t 0.0022 − 0. 0298 0.9597 0.0000 (0.2350) (0.4860) (0.0000)*** (0.8140) R d t 0.0028 0.0205 0.0237 − 0. 0002 (0.4082) (0.8904) (0.8837) (0.3851) aD. V . represents dependent v ariable. *Signi fi cant at the 10% le v el. **Signi fi cant at the 5% le v el. ***Signi fi cant at the 1% le v el. (.) represents p -v alue.

T able 10. Change in Granger causality results — premium and net b u y R f =_i,t λ f +0 λ f R1 f _i,t− 1 + f θ0 + θ f ·prem₁ f +t θ f ·BS₂ f t  dRi, t + ε f _i,t (15 ) R d i,t = λ d+0 λ dR1 d i,t− 1 +  dθ0 + θ d·prem1 d t+ θ d·BS2 d t− 1 f Ri, t− 1 + ε d i,t (16 ) R f =_i,t k f +0 k f R₁ f _i,t− 1 +  γ f +₀ γ f ·prem₁ f +t γ f ·BS₂ f t  dRi, t + k f 3  log p f i,t− 1 − c0 − c1 log p d i,t− 1  + ε f _i,t (17 ) R d i,t = k d+0 k dR1 d i,t− 1 +  dγ0 + γ d·prem1 d t+ γ d·BS2 d t− 1 f Ri, t− 1 + k d 3  log p d i,t− 1 − d0 − d1 log p f _i,t− 1  + ε d i,t (18 ) Cor . λ f (κ0 f )λ₀ f (κ1 f )θ₁ f (γ₀ f )θ₀ f (γ₁ f )θ₁ f (γ₂ f )κ₂ f ₃ Code D.V . a λ d(κ0 d)θ0 d(γ0 d)λ0 d(κ1 d)θ1 d(γ1 d)θ1 d(γ2 d)κ2 d 3 1102 R f t − 0. 0010 − 0. 0127 0.9597 − 1. 2026 0.0000 − 0. 0572 (0.2850) (0.6440) (0.0000)*** (0.0003)*** (0.7092) (0.0000)*** R d t − 0. 0009 0.0434 − 0. 0069 0.6120 0.0000 − 0. 0311 (0.4186) (0.4444) (0.9079) (0.0620)* (0.0331)** (0.0759)* 1227 R f t − 0. 0007 0.0812 0.8007 − 0. 5711 0.0000 (0.5067) (0.0098)*** (0.0000)*** (0.0187)** (0.9338) R d t − 0. 0005 0.1538 0.0185 − 0. 1088 − 0. 0000 (0.6745) (0.0050)*** (0.7185) (0.6547) (0.4754) 1402 R f t 0.0007 0.0646 0.7902 − 0. 2604 − 0. 0000 (0.8282) (0.3322) (0.0000)*** (0.7436) (0.3527) R d t 0.0015 − 0. 2796 0.0036 1.6147 − 0. 0000 (0.6548) (0.1191) (0.9751) (0.0078)*** (0.2178) 1605 R f t 0.0040 − 0. 0284 0.9316 − 0. 0000 − 0. 0001 − 0. 0007 (0.6294) (0.1631) (0.0000)*** (0.9572) (0.1677) (0.6388) R d t 0.0188 0.1373 − 0. 1393 − 0. 8032 0.0000 − 0. 0064 (0.2598) (0.1000) (0.0930)* (0.2623) (0.8310) (0.2751)

2002 R f t − 0. 0000 0.0303 1.0918 − 0. 0001 0.0002 − 0. 0546 (0.9515) (0.2814) (0.0000)*** (0.4759) (0.2958) (0.0000)*** R d t 0.0000 0.0895 − 0. 1388 0.3041 − 0. 0000 − 0. 0097 (0.9511) (0.0673)* (0.0209)** (0.2092) (0.2970) (0.5761) 2306 R f t − 0. 0002 − 0. 0065 1.0376 0.0000 0.0001 − 0. 5860 (0.7911) (0.7214) (0.0000)*** (0.5461) (0.1920) (0.0000)*** R d t 0.0001 − 0. 0385 0.0293 3.1092 − 0. 0001 − 0. 3908 (0.9193) (0.6481) (0.7429) (0.0088)*** (0.2745) (0.0000)** 2311 R f t 0.0002 0.0306 0.9584 0.0001 − 0. 0001 − 0. 0429 (0.8434) (0.2725) (0.0000)*** (0.6022) (0.3659) (0.0004)*** R d t 0.0004 0.0818 − 0. 0944 − 0. 0053 0.0000 − 0. 0437 (0.7681) (0.1888) (0.1068) (0.9800) (0.7948) (0.0029)** 2317 R f t 0.0005 0.1759 0.9618 − 0. 0000 − 0. 0005 − 0. 0510 (0.8442) (0.0069)*** (0.0000)*** (0.8481) (0.0115)** (0.0887)* R d t 0.0018 0.0943 0.0491 − 0. 1790 0.0002 − 0. 1148 (0.4687) (0.5845) (0.6458) (0.6719) (0.0711)* (0.0058)*** 2324 R f t 0.0001 0.0778 0.9645 0.0004 0.0004 (0.9716) (0.1024) (0.0000)*** (0.0955)* (0.0061)*** R d t − 0. 0005 0.2481 − 0. 2472 − 1. 5602 − 0. 0003 (0.8745) (0.1888) (0.1871) (0.3881) (0.1804) 2325 R f t 0.0004 − 0. 0396 1.0009 − 0. 0004 − 0. 0003 (0.7704) (0.2079) (0.0000)*** (0.0054)*** (0.0249)** R d t 0.0004 0.0845 − 0. 0419 − 0. 1129 0.0000 (0.7871) (0.1294) (0.4192) (0.5554) (0.7097) (continued )

T able 10. (Continued ) Cor . λ f (κ0 f )λ₀ f (κ1 f )θ₁ f (γ₀ f )θ₀ f (γ₁ f )θ₁ f (γ₂ f )κ₂ f ₃ Code D.V . a λ d(κ0 d)θ0 d(γ0 d)λ0 d(κ1 d)θ1 d(γ1 d)θ1 d(γ2 d)κ2 d 3 2330 R f _t 0.0008 − 0. 1209 0.6981 0.0002 − 0. 0004 − 0. 0198 (0.6028) (0.0016)*** (0.0000)*** (0.3534) (0.0738)* (0.1496) R d t 0.0015 0.2390 − 0. 0834 − 0. 0916 − 0. 0004 − 0. 0595 (0.1838) (0.0000)*** (0.0465)** (0.3759) (0.0000)*** (0.0000)*** 2337 R f _t 0.0000 − 0. 0153 0.7971 0.0002 − 0. 0001 − 0. 5730 (0.9963) (0.6006) (0.0000)*** (0.0745)* (0.4442) (0.0000)*** R d t − 0. 0055 0.0428 0.0331 1.3586 − 0. 0000 − 0. 1723 (0.0463)** (0.4607) (0.6081) (0.0908)* (0.9997) (0.0060)*** 2342 R f _t 0.0004 0.0019 1.0138 − 0. 0002 0.0001 − 0. 2643 (0.7814) (0.9568) (0.0000)*** (0.2635) (0.5541) (0.0000)*** R d t 0.0053 − 0. 0792 0.1627 1.4015 0.0001 − 0. 2728 (0.0954)* (0.6580) (0.3775) (0.4712) (0.7234) (0.0465)** 2344 R f _t 0.0003 − 0. 0474 1.0042 0.0000 0.0001 − 0. 1642 (0.8220) (0.1271) (0.0000)*** (0.7198) (0.4330) (0.0000)*** R d t 0.0008 0.0957 − 0. 1383 − 0. 6757 0.0002 − 0. 0649 (0.8010) (0.3544) (0.2142) (0.5575) (0.4066) (0.2318) 2345 R f _t − 0. 0008 − 0. 0755 1.0117 0.0332 0.0000 (0.3911) (0.0001)*** (0.0000)*** (0.7490) (0.2762) R d t − 0. 0001 0.0023 0.0989 − 0. 0543 0.0001 (0.9606) (0.9783) (0.2355) (0.7492) (0.0226)**