Global Finance Journal 20 (2009) 220234

Contents lists available at ScienceDirect

Global Finance Journal

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / g f j

Single-stock futures: Evidence from the Indian securities market

Umesh Kumar a, Yiuman Tse b,
a b

SUNY Canton, School of Business & Liberal Arts, Canton, NY 13617, USA University of Texas at San Antonio, Department of Finance, San Antonio, TX 78249, USA

a r t i c l e

i n f o

a b s t r a c t
Although single-stock futures (SSFs) are useful multi-purpose stock derivatives, they have not received much attention in developed markets. We analyze SSFs in the Indian market to understand their contribution in price leadership. The ndings indicate that trades in the stock market contribute more to price discovery than trades in the SSF market (72% and 28%, respectively), while quotes in the SSF market are more price innovative than quotes in the stock market (39% and 61%, respectively). Our analysis suggests that while stock and SSF trade returns have predictive ability for each other, in the case of quotes, only SSF quotes have predictive ability for stock and SSF returns. 2009 Elsevier Inc. All rights reserved.

Article history: Received 27 January 2007 Accepted 4 June 2009 Available online 31 August 2009 JEL classication: G11 G14 Keywords: Single-stock futures Price discovery Information share

1. Introduction Single-stock futures (SSFs) represent a signicant development in stock-related derivatives. It is of academic interest as to why SSFs, as a derivative product, have not gained widespread acceptance in most markets, particularly in developed markets. We analyze the Indian securities market for evidence about the role of SSFs and their effectiveness in terms of price discovery, information share, and stock returns. SSFs traded on the National Stock Exchange of India (NSE) have grown substantially since their inception in 2001. As to why other markets have struggled to generate interest among investors for SSFs, see, Fung and Tse (2008), among others, for a review of SSFs traded in a number of international exchanges. A single-stock futures contract provides a way to take advantage of arbitrage, speculative, and hedging opportunities, while reducing trading pressures on the underlying markets. Without futures contracts on individual stocks, arbitrageurs and investors must trade in the underlying assets, or trade options and index products.

Corresponding author. Department of Finance, College of Business, University of Texas at San Antonio, San Antonio, TX 78249, USA. Tel.: +1 210 458 2503; fax: +1 210 4580 2515. E-mail address: yiuman.tse@utsa.edu (Y. Tse). 1044-0283/$ see front matter 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.gfj.2009.06.004

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

221

Manaster and Rendleman (1982) show that stock option prices contain additional information compared to stock prices at the closing price level. This difference occurs due to the trading costs, the absence of tick rule governing short sales of options, and the lower margin requirements of options. The SSF has similar characteristics in terms of trading opportunities. It is a linear and efcient instrument for shortselling a stock. It further enables a cleaner hedge relative to options with potential tax advantages. In this context, it becomes imperative to explore whether SSFs contain additional information in trades or quotes as do option prices. The US typically has the most vibrant markets for stocks and derivative products. Passage of the Commodity Futures and Modernization Act of 2000 made SSFs legal in the US by repealing the Shad Johnson Accord. On November 8, 2002, two exchanges, OneChicago and the Nasdaq Liffe Market (NQLX), started SSF trading. Single-stock futures offer a cheap and exible way to gain equity market exposure for a wide range of purposes, such as hedging, speculation, and nancial engineering, yet SSF trading is relatively infrequent in the US, the largest and the most sophisticated securities market in the world. Research so far has concentrated on developed and mature markets for SSF trading. We look at the Indian market, where we nd remarkable progress in SSF trading. Since their launch in November 2001, SSFs have shown incredible progress, making the NSE the most vibrant SSF market in the world. In 2004, the NSE traded more than 25 million SSF contracts. SSF trading is also successful in the Johannesburg Stock Exchange and Eurex. Markets in the developed economies such as Sweden, Denmark, Spain, Italy, Greece, Australia, USA, UK, Euronext.liffe, Hong Kong, and Bulgaria have SSF trading but that trading is not signicant. The Futures Industry Association (July/August 2006) reports the NSE as the 13th-largest derivatives exchange by volume, and the NSE has the largest trading volume in SSFs worldwide, making it the largest global exchange for single-stock futures. In the rst four months of 2006, the worldwide volume in SSFs was 84.5 million contracts. India's NSE, the largest global exchange for single stock futures, contributed 35.4 million contracts in total volume. Studies show that SSF trading improves market efciency. Ang and Cheng (2005) nd that SSFs have a stabilizing inuence on a market. SSFs, with lower trading costs and higher leverage provide better relief for arbitrageurs than for speculators. In a study of stock futures trading in Australia, Lee and Tong (1998) conclude that SSF trading offers many of the benets associated with derivatives trading without increasing volatility or instability in the market. A coincidental increase in volume in the underlying stock market has made stock brokers less wary of losing market share and prots to the SSF market. Our research investigates the success of SSFs in the Indian market and analyzes price discovery mechanics. We examine the most comprehensive sample of stocks and stock futures available over a 12 month period (252 trading days). We also examine the information linkage between the SSFs and their stocks. Our ndings suggest that trades in the stock market perform better in terms of price discovery and information share than do trades in the SSF market. This result is contrary to previous ndings asserting that derivative products account for more price discovery and information share. However, both the stock and SSF trade returns have predictive ability for each other. Our quote analysis suggests that quotes on the SSF market lead quotes from the stock market in contributing to price discovery, and that SSF quotes have predictive ability for both stock and SSF trade returns. Both markets are mutually dependent and neither market simply free-rides on the other. More than 93% of the contracts traded in the SSF market are for single-contract trades.1 This suggests the healthy participation of retail investors in the SSF market. One plausible reason for the success of SSFs on the NSE could be the absence of an efcient or active stock-lending mechanism in the equity market. A competing hypothesis is that the rst two of the three markets (the equity market and the stock-lending market) appear to act as hidden markets for the third the SSF market. In the case of India, the equity and SSF markets came rst, and so the SSF market may be seen as a supplement to the stock-lending market. In the case of the US, the equity and the stock-lending markets developed rst, so together they act as a complement to the SSF market. This hypothesis further theorizes that even if the third market is introduced later on, it will not necessarily develop or, expand as the other two markets would continue to offer a hidden market. That may explain the lackluster response to SSFs in the US or other developed markets that have vibrant stock-lending markets.

Sometimes, institutional investors can place single-contract trades to conceal their identities.

222

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

The remainder of this study is organized as follows: in Section 2, we discuss the development of the SSF market, Section 3 describes the data construction and methodology, Section 4 contains the empirical results, and the nal section concludes the paper. 2. Development of the single-stock futures market Single-stock futures can be used as a substitute for equities for investment or speculation, as a leveraging instrument for hedging or speculation, or as a tool for price discovery in underlying stocks. Because they are inexpensive to trade, they have considerable appeal to retail investors as a way to manage their stock portfolios.2 SSFs have a downside, as their leverage can amplify any losses, and they do not provide shareholder rights. Formalized exchange trading of SSFs started in the late 1980s in Sweden, and they have recently become legal in the U.S. Now, more than 20 exchanges offer SSF products worldwide, but, by most estimates, the volume of SSF trading remains at less than 1% of total nancial-derivatives trading. SSFs have shown sluggish growth in most exchanges, even though they are sound and useful instruments, are well regulated, are offered by renowned exchanges, and provide exible instruments to achieve cost-effective hedging and portfolio rebalancing. Consequently, the instrument has not altered the dynamics of equity investing. 2.1. SSF market in the US Studies demonstrate that derivative products such as SSFs boost the trading volume in the underlying assets, enhance their liquidity, and make the whole market more efcient. The average daily turnover of SSFs in the U.S. is around 10,000 contracts. It constitutes only about 1% of the market for futures linked to the Standard & Poor's 500 stock index. This size of turnover is insufcient for a critical level of liquidity that is essential to narrowing bid-ask spreads. Recently institutional investors and other sophisticated traders have shown enthusiasm in the SSF market, but retail investors are wary and circumspect in dealing with single-stock futures. Commenting on this poor response from retail investors, Jones and Brooks (2005) state that single-stock futures prices in the US often have little relation to the prices of their underlying stocks. Their ndings imply that many hedging or large speculative trades may be difcult to execute in the current SSF market. Perhaps this situation makes institutional investors reluctant to utilize this medium. It is pertinent to understand whether the SSF market has anything to do with the bias shown by investors due to unfamiliarity with the products, or the long side of the stock market, or some other considerations. It is also important to know whether the challenges facing SSF market are due just to investors' indifference or the result of some form of regulatory initiatives. 2.2. Market design and structure in India India has a modern securities market, with 5600 rms listed on the two major stock exchanges. The exchanges are electronic and they have a T + 2 rolling settlement system. The National Stock Exchange (NSE) is the largest stock exchange in India. It is the 3rd-largest stock exchange in the world in terms of number of trades, after the NYSE and Nasdaq. Measured by the number of futures and options traded in 2004, NSE ranked as the 17th-largest derivatives exchange in the world, and the 10th-largest futures exchange. It contributes to almost all derivatives transactions in India. The value of equity derivatives trading is more than two times the value of equity trading on the NSE. The NSE has three market segments (Wholesale Debt Market (WDM) segment, Capital Market (Equity) segment, and Derivatives segment). The derivatives trading system provides fully-automated, screenbased trading for all kind of derivative products. It supports an anonymous order-driven market, which operates on a strict price time priority. Trading terminals for the derivative segment are available in more than 300 cities across the country, and trading can be accomplished by investors through the Internet.

As the SSFs are linear pay-off products, even retail investors can estimate the proceeds based on their risk appetites.

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

223

India introduced SSFs on November 9, 2001. Prior to June 2001, there was no trading of derivatives of any kind, and trading of equities was done on an accounting period settlement basis. Accounting-period trading was akin to weekly futures for the equities. However, in accounting-period trading, trades deal with a physically-deliverable asset (unlike stock or index futures, which are notional). The trades remain outstanding and are settled by actual deliveries on the settlement date. When SSFs were introduced, market participants were doubtful of their success in India, because even the U.S. did not have SSF trading. In India, retail investors are dominant in SSF trading, including the proprietary trading of small brokerage houses. The top ten member rms account for just 21% of total SSF trading, a sharp contrast to most mature markets where the top ten member rms might account for more than 60%.3 Only a small segment of SSF trading is institutional, and of that small amount, almost all comes from foreign institutional investors, who use SSF trades to carry out their hedging and portfolio-rebalancing activities. The lack of institutional trading can affect the information content in SSF trades because institutional investors are more sophisticated and have greater resources.4 A pricepressure hypothesis implies that institutional trades inuence price formation in a market more than do trades by individuals. Despite a short history of derivative trading in India, in the rst two months of 2005 NSE conducted 35 times more trading in SSF contracts than did OneChicago. This paper looks at plausible reasons for the success of SSFs in India. The practice of badla and accounting-period trading has been credited to an extent for this success. The badla was a quasi-derivative product and conceptually close to futures contracts, since it was used to defer settlement in the equity market. Individual investors used the badla market before the introduction of the SSF market. These investors had years of practice trading something akin to equity futures. So, when the badla products ceased to be available in June 2001 and SSFs were introduced in November 2001, these investors easily migrated to single-stock futures. However, other features that have made the SSF market successful are as follows: (a) It is an order driven market having similar characteristics to the equity market, implying that individual investors familiar with the equity market system can easily understand the execution of trades. (b) Individual investors generally put 1530% of the total value of an SSF trade on margin with their broker. (c) The SSF contract size is affordable since its value is just above Rs 0.2 million (approximately USD 4500). Thus, the affordability and familiarity of the products have made it easier for individual investors to participate heavily in the SSF market.5 3. Data construction and methodology 3.1. Data source This study employs data from high-frequency stocks and their SSFs, obtained from the National Stock Exchange of India (NSE) for the period January 2004 through December 2004, a total of 252 trading days. All derivatives trading in India is overwhelmingly concentrated in the NSE. We choose only NSE trade data for stocks, since its stock market segment contributes almost 2/3 of total trading volume in India. The data are in two segments (trade data and snapshots of limit order books). The trade data contain the details of all trades that took place in the exchange for the stocks and SSFs. The snapshots of limit order book for stocks were taken at four different times during the day. In the case of SSFs, the snapshots of the limit order book were taken at ve different times. The limit order book contains all limit orders entering the NSE trading system (right to trade against them, without any obligation) and they are free options which anyone can exploit. The order book snapshot time for stocks is 11 A.M., 12 noon, 1 P.M., and 2 P.M. The order book snapshot time for SSFs is 11 A.M., 12 noon, 1 P.M., 2 P.M., and 3 P.M. The normal market

3 NSE monthly derivates update (December 2004) reports that the top ve members make 12% valuewise contribution while the next ve have 9%. 4 By examining the information content of institutional trades on the London Stock Exchange, Bozcuk and Lasfer (2005) found that the type of investors, the combination of the size of the trade, and investors' resulting level of ownership are the major determinants of price impact. Institutional and individual investors observe news or price movements in different ways, process such information differently, and thus trade accordingly. 5 Presently, the SSF market in South Africa (Johannesburg Stock Exchange) has been doing well and credit goes to the affordability and participation of retail investors; the contract price in SSF market is sufciently low to encourage participation by retail investors.

224

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

operation time for stock and SSF markets in the NSE is synchronized, with trading starting at 9:55 A.M. and closing at 3:30 P.M. The snapshots obtained are the pictures of the complete limit order book at a given point in time. The limit order book is an electronic order matching embedded in all orders present at any point in time. Though the snapshots data provide a picture at a few time points during the day, many plausible answers related to liquidity, market impact, and bid-ask spread can be obtained accurately. The snapshots are rich in terms of quote characteristics, such as order ID number, ticker symbol, quantity, price, timestamp, buy or sell, type of order (day order, good till date, cancel, immediate/cancel), quantity ags (minimum ll, all or none, disclosed quantity), quantity disclosed, and price ags (at the open price, market price, and stop loss order). We mainly use price, timestamp, and buy or sell data for our study. The buy or sell indicator allows for calculation of spread and mid quote price. Timestamp and price facilitate the construction of desired data series for stocks and their SSFs in price discovery analysis.6 The exchange selects SSF stocks from the top 500 stocks based on average daily market capitalization and daily traded value for the previous six months. We restrict our sample to only those SSFs that have daily trading volume above 1000 contracts. Based on this criterion, we initially selected 40 SSFs. These SSFs and their stocks are the most liquid and actively-traded securities on the NSE. We found that some of these rms have merged, changed their name or split, and sometimes their trading volume became too low to match our criterion. We therefore eliminated such SSFs from our sample. In other cases where we could obtain the desired data series from the raw data we were forced to drop those SSFs from our nal sample. The resulting sample for our study is comprised of 30 stocks and their SSFs. These 30 SSF contracts contribute 8085% of total trading volume. Similarly, their stocks represent 8590% of total stock trading volume. These stocks are all constituents of the primary index of the exchange (the S&P Nifty Index). Our dataset is more comprehensive and larger than the datasets used in many previous studies, and the integrity of the data is strong, since the data are obtained directly from the exchange. 3.2. Data preparation The trade data contain the details of all trades which occurred at the exchange for the stocks and SSFs on a daily basis. All trades are time stamped. We sort the trade data based on the time stamp and choose the last trade from each minute of trades. The quote data are obtained from the limit order book snapshot that lists all outstanding orders of all securities at the time the snapshot is recorded. The orders are timestamped and identied as buy or sell orders. We merge these snapshots on a daily basis for each stock separately. The merged snapshot data are sorted based on the time stamp. We follow this procedure for stock and SSF quotes separately. Order Type indicates buy (B) or sell (S) orders. From these daily les, we select the highest bid quote and lowest ask quote for every minute for both stocks and SSFs. We calculate the mid-quote after averaging the bid and ask quotes for each minute. Hence the data generated are similar for both stocks and SSFs. We omit the outliers, if any, from trade and quotes to avoid contamination of the data series. A trade price is an outlier, if the SSF price is 5% above or below the stock price. A quote price is treated as an outlier, if the SSF midquote is 5% above or below the stock midquote. Both lters remove less than 3% of all observations from the nal data. Following this procedure, we obtain 2,429,656 and 1,648,650 minute-to-minute observations for trades and quote les, respectively. 3.3. Understanding regulatory implications and institutional trading One of the tasks of this paper is to develop an understanding of regulatory initiatives toward SSFs. We pay particular attention to July 2004, since there was a regulatory intervention in that month. New regulations increased the exposure and position limits for Foreign Institutional Investors (FIIs) in derivative products. We scrutinize whether such regulatory changes facilitate SSF trading by impacting price discovery activities. Institutional investors, including FIIs, were allowed to trade in SSFs in 2002, but faced a restrictive burden in terms of limits to position and open interest. In July 2004, the capital market regulator eased some restrictions on FIIs for position limits on derivative products, including doubling the market-wide position limit for SSFs.
6

See Chakrabarty and Jain (2005) for a detailed description of the data.

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

225

3.4. Price discovery and information share between the SSF and stock markets One implicit assumption in the studies of price discovery is that markets share at least one common driving force, leading to the use of common-factor models in such studies. These models derive their results from the same economic rationales. To investigate price discovery and information share between and SSF and stock markets, we use two popular common-factor models, from Hasbrouck (1995) and Gonzalo & Granger (1995). Baillie, Booth, Tse & Zabotina et al. (2002), De Jong (2002), and Lehmann (2002) establish the relationship between these two models. Harris, McInish, & Wood, (2002) and Hasbrouck (2002) enumerate the differences between the two models. Booth, Lin, Ji-Chai, Martikainen & Tse, (2002) use both models to examine the Finnish upstairs and downstairs stock markets. A number of other studies have also used the information-share and/or permanent-transitory models (see, e.g., Brockman & Tse, 1995; Ding, Harris, deB, Lau & McInish, 1999l; Tse & Erenburg, 2003). Both models use the vector error correction model (VECM) as their basis, but they differ in their price discovery mechanisms. The Hasbrouck (1995) model denes price discovery in terms of the variance of the innovations to the common factor, and measures each market's relative contribution to this variance. This contribution is called the market's information share. The Gonzalo & Granger (1995) model, however, focuses on the error correction process and the components of the common factor. This process involves only permanent (as opposed to transitory) shocks that result in a disequilibrium. The Gonzalo and Granger model measures each market's contribution to the common factor, where contribution is dened as a function of the market's error correction coefcients. The feature that distinguishes the models from each other is that the Hasbrouck (1995) model decomposes the variance of the implicit efcient price. Relying on the premise that price volatility reects the ow of information, it attributes a greater share of efcient price discovery to the market that contributes the greatest share to this volatility. By contrast, the Gonzalo & Granger (1995) model approach decomposes the common factor itself. In doing so, the Gonzalo and Granger model ignores the correlation among the markets and attributes the leading role solely to the market that adjusts least to the price movements in other markets. In markets affected by the same information ow (i.e. with similar volatility), these two models produce consistent results, i.e. a market with the greatest contribution to the price discovery has the largest loading on the common factor. Both information-share and permanent-transitory models are derived from a vector error correction model (VECM) in the following form: Xt = Xt1 + i Xt1 + t
i=1 k

Where Xt = {Xit} is an n 1 vector of cointegrated prices. and s are n n matrices of parameters, and t is an n 1 vector of serially-uncorrelated residuals with a covariancecovariance matrix = {ij}. The long run relation matrix has a reduced rank of r < n and can be decomposed as = , where and are n r matrices. The matrix consists of the cointegrating vectors and is the error correction (or equilibrium adjustment) matrix. If r = n 1 and is spanned by the differentials of each pair of price series, then all Xit are driven by one common factor. This is the case for stock and stock futures prices. Hasbrouck (1995) transforms the VECM into an integrated form of a vector moving average (VMA): Xt = J + *Lt
t =1 k

where J(1,,1) is a column vector of ones, = (1,, n) is a row vector, and * is a matrix of polynomials in the lag operator, L. The Hasbrouck (1995) model denes a market's contribution to price discovery as its information share the market's proportion of the variance of the efcient price innovation. By contrast, the Gonzalo and Granger (1995) model decomposes the common factor into a linear combination of the prices. An advantage of the Gonzalo and Granger model is that the common-factor estimates are exactly identied, as they do not depend on the ordering of the variables.

226

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

Baillie et al. (2002) and De Jong (2002) show that the information-share and permanent-transitory models provide similar results if the contemporaneous cross-equation residuals are uncorrelated. If there is a strong correlation among the contemporaneous cross-equation residuals, differences in the results from the two models can be substantial. Hasbrouck (1995) points out that the information share estimates will depend on the ordering of variables in the Cholesky factorization, if the price innovations are correlated. Martens, Kofman and Vorst (1998), Baillie et al. (2002), Booth et al. (2002), and Huang (2002) also report a substantial difference in their Hasbrouck upper and lower bounds of information shares. For a bivariate case, Baillie et al. (2002) and Tse (1999) show that the average of the information shares given by two permutations is a reasonable estimate of a market's role in price discovery. We use the average of information shares to interpret the results. 3.5. Information linkage between the SSF and stock markets Following Chan, Chung, and Fong, (2002), we use a multivariate VAR model to analyze the informational role and interdependency between SSFs and their stocks.7 The model also investigates the dynamic relationship among trades and quote revisions for the SSFs and their stocks. The following VAR model is used in the regression. rt = a1 rt1 + + ap rtp + bo zt + b1 zt1 + + bp ztp + 1;t zt = c1 rt1 + + cp rtp + d1 zt1 + + dp ztp + z;t 3 4

where is rt the trade return for stocks and SSFs at transaction time t, and zt is the quote return for stocks and SSFs at transaction time t, dened as the change in midquote from the quote following transaction t 1 to the quote following transaction time t. We use minute-to-minute trade and quote prices to calculate the returns. We calculate the mean and standard deviation for these return variables for each security on a daily basis. Following Easley, O'Hara, and Srinivas (1998), the return variables for each security are standardized by subtracting the mean and then dividing by the standard deviation. The standardized returns help to control for cross-sectional variations across different SSFs and stocks. We choose six lags for each explanatory variable in the regression. In the trade returns regressions, we use separate lagged stock and SSF returns as explanatory variables separately to eliminate the possibility of multicollinearity. The leadlag relationship between trade and quote returns should demonstrate the informational effects among them. When the leadlag relationships of returns are used in more than one market setup, they explain the informational role and their linkages between the individual markets. If informed investors submit limit orders in the stock market only, SSF quotes will not show predictive ability for stock returns. If informed investors submit limit orders to the SSF market to disguise their private information and exploit it by transacting in the stock market, the SSF quotes will have predictive ability for both stock and SSF returns. If the markets are dependent on each other, the returns on SSFs and stocks will have predictive ability for each other. 4. Empirical results 4.1. Relationship between the stock and SSF markets Fig. 1 illustrates monthly trading turnover of all sample stocks and SSFs. We nd that the turnover of SSFs is higher than the turnover of stocks, except in the months of May and June, 2004. Overall, we nd that SSFs have substantial trading, almost 1.6 times the number of trades of stocks. Trading volume is highest for both stocks and SSFs in January, 2004. Trading volume gradually slows down in subsequent months, and reaches its lowest level in June, 2004. Later, the trading volume gains in both segments.
7 Chan et al. (2002) analyze the informational linkage between the option and stock markets by integrating quote returns and net trade volume to see the predictive power for both option and stock returns.

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

227

Fig. 1. Monthly trading turnover in stock vis--vis SSF. We calculate monthly stock and SSF turnover. The straight line indicates the monthly stock turnover, while the dotted line signies the SSF turnover. The turnover is denoted in Indian currency in billion rupees.

4.2. Trade size in the SSF market First, we calculate the average number of monthly trades, number of contracts per trade, percentage of trades, and percentage of volume in the SSF market. We try to understand the kind of investors dominating the SSF market. Table 1 exhibits our ndings. We expect institutional investors to make use of the advantages offered by SSF trading. We examine this assumption using size of trades transacted in the SSF market. We nd that single contract trades overwhelmingly dominate the SSF market. On average, single contract trades account for more than 93% of all contracts traded. Two contract trades constitute only 4.35% of total trades, while trades in three or more contracts comprise only 2.47% of total trades. The notional value of a single contract size is comparatively small (approximately USD 4500). Trade size clearly suggests that institutional trades do not rule the SSF market. In this market, we assume that institutional investors would tend to deal in more than a single contract trades, considering the transaction and other attendant

Table 1 Trade size in SSFs. Month Jan-04 Feb-04 Mar-04 Apr-04 May-04 Jun-04 Jul-04 Aug-04 Sep-04 Oct-04 Nov-04 Dec-04 Average One-contract ('000) 2330 1712 1680 1980 1641 1691 1785 1807 1845 1858 1768 2552 1887 Two-contract ('000) 113 68 319 132 113 100 110 115 118 117 113 168 132 Three-or- more contract ('000) 60 37 169 81 72 63 69 74 82 80 80 114 82 Total contract ('000) 2503 1818 2168 2193 1827 1855 1965 1997 2046 2055 1962 2835 2102 % of one contract 95.17 96.22 83.73 93.44 92.07 94.19 94.67 94.35 93.91 93.91 93.48 93.05 93.18 % of two contract 3.23 2.48 11.89 4.21 3.94 3.33 3.38 3.55 3.85 3.87 4.05 4.43 4.35 % of three or more contract 1.60 1.30 4.38 2.35 3.98 2.48 1.95 2.10 2.24 2.21 2.47 2.52 2.47

We calculate trade sizes for all SSF transactions. We derive monthly contracts in the SSF market from daily trade data. The trade size is segmented into three parts i.e. one-contract trades, two-contract trades, and three-or-more contract trades. From monthly data, we compute the monthly average trade size in the SSFs.

228

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

Table 2 Percentage spread in stock and SSF quotes. Panel A Monthly percentage spread of stock and SSF quotes Month N Mean Stock Jan-04 Feb-04 Mar-04 Apr-04 May-04 Jun-04 Jul-04 Aug-04 Sep-04 Oct-04 Nov-04 Dec-04 Panel B Mean percentage spread of stock and SSF quotes Variable Stock SSF Difference N 252 252 252 Mean 1.63 2.19 0.56 Standard deviation 0.65 0.82 0.35 Standard error 0.04 0.05 0.02 t-statistic 39.80 44.49 25.29 p-value <0.0001 <0.0001 <0.0001 21 19 22 20 21 22 22 21 22 20 19 23 2.21 1.76 1.69 1.60 2.60 1.61 1.54 1.29 1.19 1.35 1.34 1.40 SSF 2.68 2.49 2.31 2.12 3.51 2.69 2.50 1.85 1.49 1.64 1.51 1.51 Difference 0.48 0.73 0.62 0.52 0.91 1.08 0.96 0.56 0.30 0.29 0.16 0.11 Standard deviation Stock 0.54 0.34 0.18 0.19 1.66 0.18 0.30 0.09 0.08 0.13 0.13 0.10 SSF 0.53 0.43 0.18 0.13 1.53 0.27 0.36 0.15 0.12 0.15 0.15 0.16 Difference 0.23 0.23 0.16 0.16 0.29 0.17 0.20 0.14 0.11 0.13 0.12 0.16 t-statistic Stock 18.88 22.84 44.28 38.65 7.18 41.41 24.46 66.82 66.92 48.11 45.72 68.71 SSF 23.44 24.99 60.45 72.25 10.50 47.16 32.44 55.33 58.77 47.52 43.24 46.55 Difference 9.69 13.84 18.13 14.93 14.60 29.27 23.07 17.77 12.84 9.80 5.87 3.36

The statistics consist of stock and SSF quote spreads for the sample period. The sample data are segmented into two panels. The percentage spread is measured as 100% (Ask Bid) / Midquote, where the midquote is the average of the bid and ask prices. The stock and SSF quote spreads are derived from their mean spread on a daily basis. We obtain percentage spread differences by subtracting the percentage stock spread from the percentage SSF spread. The signicance of percentage spread difference from zero is tested by t-tests. Panel A represents the percentage spread in stock and SSF quotes on a monthly basis, while Panel B represents the percentage spread in stock and SSF quotes for entire period.

costs associated with doing single-contract transaction. The trading pattern above (overwhelmingly single contract trades) is consistent in almost all months. Therefore, we believe that there is a strong retail participation in the SSF market, as claimed by the exchange and other market intermediaries.8 4.3. Bid-ask spreads Table 2 presents the percentage spread for stock and SSF quotes. Percentage spread is measured as 100% x (Ask Price Bid Price) / Midquote, where midquote is the average of bid and ask prices. We nd that the mean percentage spread of stock quotes ranges between 1.19% and 2.60%. SSF quotes show mean percentage spreads between 1.49% and 3.51%. The differences between stock and SSF quotes vary between 0.11% and 1.08%. Overall, the mean percentage spread for stock quotes is 34% lower than SSF quote spreads. Similarly, volatility in stock quotes is 21% lower than the volatility in SSF quotes. The difference between the average mean spread of stock and SSFs quotes is 0.56%. Stock quotes show lower spreads than SSF quotes. It is important to note that there is a difference between quote-setting behavior in stocks and SSFs. We assume that SSFs should have lower spreads, but when we analyze trade size, we nd that SSF trading is mostly driven and inuenced by retail participation. Hence, the higher spread observed in SSF quotes is not surprising. It is interesting to note that the spreads
8 NSE monthly derivatives update (December 2004) mentions that non-institutional investors contribute more than 95% of total derivatives trading. FOW (Issue 403 dated December 01, 2004) reports that in the SSFs, the main participants are retail traders and proprietary trading by member rms, followed distantly by foreign institutional business, and domestic mutual funds. Bloomberg News (April 05, 2006) reports that retail investors account for 63% of the total trading in stock futures in the Indian market.

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234 Table 3 Information share from trades. Month Hasbrouck information share Stock Jan-04 Feb-04 Mar-04 Apr-04 May-04 Jun-04 Jul-04 Aug-04 Sep-04 Oct-04 Nov-04 Dec-04 Average 0.79 0.77 0.78 0.72 0.66 0.80 0.50 0.49 0.81 0.76 0.78 0.76 0.72 SSF 0.21 0.23 0.22 0.29 0.34 0.20 0.50 0.51 0.19 0.24 0.22 0.24 0.28 GonzaloGranger factor weights Stock 0.84 0.84 0.80 0.77 0.70 0.83 0.51 0.39 0.84 0.80 0.80 0.79 0.74

229

SSF 0.16 0.16 0.20 0.23 0.30 0.16 0.49 0.61 0.16 0.20 0.20 0.21 0.26

The table reports price discovery results based on the Hasbrouck (1995) model and the Gonzalo and Granger (1995) model for stock and SSF trades. The prices are calculated at one-minute intervals. The information share is the proportion of variance in the implicit efcient price of the stock that is attributable to innovations in that market. The panel represents the information share on a monthly basis, and we then compute the average over all months to get the overall information share for entire period.

are wider in stock and SSF quotes during June and July 2004, when the trading volume of the SSF market shrinks and becomes almost equal to that of the stock market. When relative trading volume in the SSF market increases, the spread differences narrow. 4.4. Price discovery and information share between the SSF and stock markets 4.4.1. Results from trade transactions Table 3 reports the price discovery results for trade prices of SSFs and their stocks at one-minute intervals. Both models show that the stock market produces higher price discovery in all months except July and August, 2004. In these two months, each market contributes almost equally in price discovery and transmission. We notice that the average information shares for stocks and SSFs are 0.72 and 0.28, respectively. The ndings suggest that information production and price discovery occur in the stock market. Despite higher turnover volume in the SSF market, as shown in Fig. 1, the contribution of SSFs to price discovery is modest. The Gonzalo and Granger model provides similar results; 0.74 (stocks) and 0.26 (SSFs). Thus, our result is in contrast with other ndings showing futures market to be more efcient in price discovery. However, we nd that both markets have almost equal roles in price discovery during July and August. We consider two signicant events in this period. First, there was a change in regulatory limitations for Foreign Institutional Investors (FIIs) trading in stock index futures. Second, in the case of SSFs, there was a relaxation in market-wide position limits. These two factors may have inuenced price-discovery and information-share contribution. It is worth noting that FIIs are sophisticated investors and are primary players in institutional trades for derivative products, including SSFs, while domestic institutional investors are relatively inactive in derivative products, particularly in SSFs. The relaxation of regulation in stock index futures also affects the 30 sample stocks, since they are constituents of the indexes. The FIIs benet from increases in market-wide position limits. During our sample period, they average 60% of the total open interest in the SSFs.9 Open interest is a measure of how much interest a particular product garners from investors. FIIs have a higher level of open position in the SSFs which demonstrates their level of interest. Chan and Lakonishok (1995) report that the estimates of the price impact of institutional trades are substantially higher when trades are evaluated not individually but in the broader context of a package.

Data obtained from SEBI's Annual Report 200405.

230 Table 4 Information share from quotes. Month

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

Hasbrouck information share Stock SSF 0.56 0.62 0.59 0.60 0.44 0.52 0.54 0.62 0.74 0.71 0.71 0.71 0.61

GonzaloGranger factor weights Stock 0.44 0.38 0.41 0.40 0.57 0.48 0.42 0.31 0.24 0.28 0.28 0.27 0.37 SSF 0.56 0.62 0.59 0.60 0.43 0.52 0.58 0.69 0.76 0.72 0.69 0.74 0.63

Jan-04 Feb-04 Mar-04 Apr-04 May-04 Jun-04 Jul-04 Aug-04 Sep-04 Oct-04 Nov-04 Dec-04 Average

0.44 0.39 0.41 0.41 0.56 0.48 0.46 0.39 0.26 0.29 0.29 0.29 0.39

The table reports price discovery result based on the Hasbrouck (1995) model and the Gonzalo and Granger (1995) model for stock and SSF quotes. The quotes are computed at one-minute intervals. The panel represents the information share on a monthly basis, and then we compute average to get overall information share for entire period.

Frino, Walter, and West (2000) document that investors with better market-wide information are more likely to trade in stock index futures, strengthening the price discovery role of futures market signicantly around macroeconomic news releases. Bozcuk and Lasfer (2005) nd that the type of investors behind the trades, the combination of the size of the trades, and the investors' resulting level of ownership are major determinants of the price impact. In our case, FIIs are major shareholders in the sample stocks. These ndings corroborate that, after the relaxation in position limits in July 2004, the increase in FIIs trading in the SSFs alters the information content in prices.10 The bulk of the trades in the NSE do not come from any innate complementary hedging function that SSFs offer, or even from any competitive advantage they enjoy over the stock markets. 4.4.2. Results from competitive quotes A market that is a price leader or information producer does not necessarily also provides the best quotes. It simply indicates that the market impounds information faster than the others. Moreover, SSFs may not trade at exactly the same price as their stocks, but they will trade at a price that is very close because of the well-known cost-of-carry relationship. Table 4 reports price discovery and information share results for stock and SSF quotes. To understand the quality of quotes, it is important to understand the characteristics of the markets from which these quotes originate. Price discovery inferred from quotes does not necessarily reect the market in which informed traders trade. When similar securities are traded in multiple markets, informed investors sometimes conceal their private information by submitting limit orders in the market where they do not intend to execute trades. They may submit limit orders in a market where the cost is lower to exploit their private information and subsequently cancelling these limit orders. In such a situation, the private information is rst reected in the quote revisions and not in trades. There is a separation between price leadership from quotes and price leadership from trades. These results are different from those reported in Table 3 using trades. We nd that stock and SSF quotes yield an average information share of 0.39 and 0.61, respectively, while common factor coefcients are 0.37 and 0.63, respectively. This indicates that SSF quotes lead stock quotes in price discovery.

10 The Holden and Subrahmanyam (1992) model illustrates that greater competition among strategic informed traders, such as institutional investors, results in faster incorporation of private information. Boehmer and Kelley (2005) show that increases in institutional trading volume are associated with greater informational efciency. Moreover, Dey and Radhakrishna (2001) and Fehle (2004) nd that institutional trading and institutional ownership are negatively related to bidask spread.

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

231

The above results suggest that trades on the stock market have higher information share while quotes on the SSF market contribute more in price discovery. The Hasbrouck (1995) model does not determine which market has the best prices. The information share in the model measures who moves rst in the process of price adjustment. Hasbrouck (1995) nds that information share of non-NYSE (regional exchanges and Nasdaq) quotes is negligible, about 5%. Harris et al (1995) demonstrate that non-NYSE trades provide substantial price discovery, about 30%. This suggests that non-NYSE quotes have negligible information share (5%), while the non-NYSE trades have substantially higher information share (30%). Tse (2000) further shows that non-NYSE trades provide price discovery, but non-NYSE quotes do not. This is similar to our nding that quotes on the stock market have low information share (39%), while trades on this market have high information share (72%). As discussed in Hasbrouck (1991, Hasbrouck, 1995), while off-NYSE quotes are autoquotes that simply follow the NYSE quotes, the actual mechanism linking trade and quote responses is confounded by different market-microstructure aspects, such as liquidity effects, inventory-control behavior, and price discreteness. Similarly, quotes from the stock market (like the off-NYSE autoquotes) may automatically adjust to the quotes from the SSF, resulting in a higher information share from quotes in the SSF market. The proportion of trades by domestic institutional investors in the SSF market is low. This fact may explain the lower information share for SSF trades. Additionally, the higher information share for stock trades may also arise due to the following institutional feature. In the Indian stock market, institutional investors cannot sell shares unless they possess the shares in their account. Since settlement occurs on T + 2, this requirement imposes a minimum effective holding period of 2 days. Thus, they cannot day trade, and institutional investors will place a buy order only when they are condent about the stock's price performance over the next two days. This feature will increase the information content of cash trades and is consistent with the higher proportion of information share for cash market trades. The SSFs analyzed in the study are one month contracts. Since the contract period is short, institutional investors generally do not use these SSFs for hedging. The fact that SSF trades are settled only by cash rather than by stock delivery also encourages speculative trades. Lower transaction cost in the SSFs enables liquidity providers to revise quotes more quickly, and contributes to the higher information shares of the SSF quotes. 4.5. Results from VAR model Table 5 presents the results. First, we nd that stock returns are positively related to rst-lag stock returns (0.339) and negatively related to second-lag stock returns (0.101). They are positively related to SSF trade and quote returns. However, there is a weak relationship between stock returns and their quote returns. The results indicate that lagged stock and SSF returns have predictive ability for the stock returns. Similarly, the contemporaneous and lagged SSF quote returns can predict the stock returns. The signicance level of stock and SSF quotes indicates that SSF quote returns can inuence the stock returns while stock quote returns may not. SSF quotes inuencing the prices of stocks and SSFs show that they have higher information content and that stock quotes adjust to SSF quotes. Second, we nd that SSF returns are positively related to the rst lag stock returns (0.268) and negatively related to the second-lag stock returns (0.053). Stock and SSF returns both are positively related to the rst-lag stock returns indicating the higher information ow in trades of the stock market. Further, we nd that the contemporaneous and lagged SSF quote returns are positively related to stock and SSF returns indicating the higher information production in the quote revisions. As expected, we nd that quote returns for stock and SSF are negatively related to their own lag returns, respectively. Overall, the regression results suggest that stock and SSF trade returns have predictive ability for each other. In case of quotes, only SSF quotes have predictive ability for both stock and SSF returns. This corroborates the ndings in Tables 3 and 4 that trades in the stock market are more informative, while quotes in the SSF market have more information content. These results are consistent with the information transmission between the two markets. This indicates that the markets are mutually dependent and neither market simply free-rides on the other market. As a robustness check, we use standardized veminute returns, raw one-minute returns, and raw ve-minute returns in the VAR model. However, the results remain qualitatively similar.

232

Table 5 Regression analysis of the relationship between standardized one-minute trade and quote returns of stocks and their SSFs. Explanatory variables U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234 Lagged stock return Dependent variable Stock returns Stock returns SSF returns SSF returns Stock quote returns SSF quote returns 0.001
(0.61)

Lagged SSF returns Lag 1 Lag 2 Lag 0 0.002

(1.98)

Lagged stock quote returns Lag 1 0.002

(2.37)

Lagged SSF quote returns Lag 2 0.004

(3.95)

Lag 1 0.339
(372.79)

Lag 2 0.101
(105.32)

Lag 0 0.009
(10.48)

Lag 1 0.024
(24.37)

Lag 2 0.026
(25.90)

0.261
(320.50)

0.051
(62.49)

0.002
(2.25)

0.001
(1.22)

0.004
(3.79)

0.009
(10.48)

0.025
(25.48)

0.028
(27.56)

0.268
(267.27)

0.053
( 50.40)

0.001
(0.68)

0.001
(0.88)

0.001
(0.54)

0.008
(8.38)

0.020
(18.80)

0.019
(16.87)

0.052
(56.78)

0.000
(0.30)

0.001
(0.53)

0.000
(0.40)

0.001
(0.74)

0.008
(8.19)

0.023
(21.37)

0.026
(23.16)

0.001
(1.17)

0.000
(0.44)

0.001
(0.63)

0.482
(526.11)

0.277
(272.79)

0.003
(3.18)

0.003
(3.14)

0.000
(0.07)

0.001
(1.06)

0.001
(1.22)

0.003
(2.29)

0.002
(2.69)

0.004
(3.47)

0.448
(488.60)

0.294
( 293.93)

This table presents the results of the following multivariate VAR model: rt = a1 rt1 + + ap rtp + bo zt + bo zt1 + + bp ztp + 1;t zt = c1 rt1 + + cp rtp + d1 zt1 + + dp ztp + 2;t Where rt is the trade return for stocks and SSFs at transaction time t, and zt is the quote return for stocks and SSFs at transaction time t, which is the change in midquote from the quotes following transaction t 1 to the quotes following transaction time t. The return variables are standardized by subtracting the mean and then dividing by the standard deviation. We report the regression coefcients for the contemporaneous and rst two lags and t-statistics in parenthesis.

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

233

5. Conclusions Single-stock futures (SSF) are a puzzling derivative product. They are useful multi-purpose products, but have not gained market share in developed countries. In contrast, SSFs have done well in the Indian securities market. Hence, we study the Indian SSF market to understand its characteristics and price discovery process between the SSFs and their underlying stocks. Our paper provides useful insights into the success of SSFs in the Indian market. We nd that the stock market performs better in terms of price discovery and information share for trades. This result is contrary to the evidence from other studies, where derivative products enjoy more price discovery and information share. However, we nd that SSF quotes lead stock quotes in price discovery contribution. This means that SSF quotes are better and more informative than stock quotes. We nd that stock and SSF trade returns have predictive ability for each other. The quote analysis indicates that SSF quotes have predictive ability for both stock and SSF returns. Hence, the markets are mutually dependent and neither market simply free-rides on the other. Overall, our ndings suggest that a vibrant SSF market requires affordable contracts and retail investors' participation. Regulatory initiative is also a vital component in the expansion of the SSF market. Acknowledgements Tse acknowledges the nancial support from a summer research grant of U.S. Global Investors, Inc. and the College of Business at The University of Texas at San Antonio. We would like to thank John Wald and James Hackard for their useful suggestions. Appendix A

Contract specications for single-stock futures in the NSE Contract size Tick size Trading cycle As specied by the exchange subject to a minimum value of Rs. 0.2 million Rs 0.05 A maximum of three month trading cycle the near month (one), the next month (two), and the far month (three). New contract is introduced on the next trading day following the expiry of near month contract Upfront initial margin on a daily basis Last Thursday of the expiry month or the preceding trading day, if the last Thursday is a holiday Operating range of 20% of the base price Last trading day In cash on T + 1 basis Closing price of futures contract on the trading day Closing value of underlying security on the last trading day of the futures contract

Margins Expiration day Price band Settlement day Settlement Daily settlement price Final settlement price Source: www.nseindia.com.

The contract specications for stock futures and options are similar. The SSF has contracts with 1 month, 2 months, and 3 months to expiry at any point of time. In case of stock options, there are a minimum of 7 strike prices, three in-the-money, one at-the-money and three out-of-the-money for every call and put option (for American options only). The last Thursday of the respective expiry month is the expiry day and a new contract is introduced on the next trading day of the near month contract. All derivatives products in India including SSFs and options are cash settled. The contract size is a minimum value of Rs. 0.2 million. There is an upfront initial margin for the trades. The mark to market margin is levied at the daily settlement price. The risk management system for SSFs and stocks is separately performed and no cross margining is available. References
Ang, J. S., & Cheng, Y. (2005). Financial innovations and market efciency: The case for single stock futures. Journal of Applied Finance, 15(1), 3851. Baillie, R. T., Booth, G. G., Tse, Y., & Zabotina, T. (2002). Price discovery and common factor models. Journal of Financial Markets, 5, 309321.

234

U. Kumar, Y. Tse / Global Finance Journal 20 (2009) 220234

Booth, G. G., Lin, Ji-Chai, Martikainen, T., & Tse, Y. (2002). Trading and pricing in upstairs and downstairs stock markets. The Review of Financial Studies, 15, 11111135. Boehmer, E., & Kelley, E. (2005). Institutional investors and the informational efciency of prices. Working Paper, Texas A&M University. Bozcuk, A., & Lasfer, M. A. (2005). The information content of institutional trades on the London Stock Exchange. Journal of Financial and Quantitative Analysis, 40(3), 621644. Brockman, P., & Tse, Y. (1995). Information shares in Canadian agricultural cash and futures markets. Applied Economics Letters, 2, 335338. Chan, K., Chung, P. P., & Fong, W. (2002). The informational role of stock and option volume. The Review of Financial Studies, 15, 10491075. Chan, L. K. C., & Lakonishok, J. (1995). The behavior of stock prices around institutional trades. The Journal of Finance, 50, 11471174. Chakrabarty, B., & Jain, J. (2005). Understanding the microstructure in Indian markets. National Stock Exchange of India Research Paper Series #128. De Jong, F. (2002). Measures of contributions to price discovery: A comparison. Journal of Financial Markets, 5, 323328. Dey, M. K., & Radhakrishna, B. (2001). Institutional trading, trading volume, and spread : NBER. Ding, D. K., Harris, F. H., deB, Lau, S. T., & McInish, T. H. (1999). An investigation of price discovery in informationally-linked markets: Equity trading in Malaysia and Singapore. Journal of Multinational Financial Management, 9, 317329. Easley, D., O'Hara, M., & Srinivas, P. (1998). Option volume and stock prices: Evidence on where informed traders trade. The Journal of Finance, 53, 431465. Fehle, F. (2004). Bidask spreads and institutional ownership. Review of Quantitative Finance and Accounting, 22, 275292. Frino, A., Walter, T., & West, A. (2000). The leadlag relationship between equities and stock index futures markets around information releases. Journal of Futures Markets, 20, 467487. Fung, J. K. W., & Tse, T. (2008). Efciency of single-stock futures: An intraday analysis. Journal of Futures Markets, 28, 518536. Gonzalo, J., & Granger, C. (1995). Estimation of common long-memory components in cointegrated systems. Journal of Business & Economic Statistics, 13, 2736. Harris, F. B., McInish, T. H., & Wood, R. A. (2002). Security price adjustment across exchanges: An investigation of common factor components for Dow stocks. Journal of Financial Markets, 5, 277308. Harris, F. B., McInish, T. H., Shoesmith, G. L., & Wood, R. A. (1995). Cointegration, error correction, and price discovery on informationally linked security markets. Journal of Financial and Quantitative Analysis, 30, 563.579. Hasbrouck, J. (1995). One security, many markets: Determining the contributions to price discovery. The Journal of Finance, 50, 11751199. Hasbrouck, J. (2002). Stalking the efcient price in market microstructure specications: An overview. Journal of Financial Markets, 5, 329336. Holden, C. W., & Subrahmanyam, A. (1992). Long-lived private information and imperfect competition. The Journal of Finance, 47, 247270. Huang, R. D. (2002). The quality of ECN and Nasdaq market makers quotes. The Journal of Finance, 57, 12851319. Jones, T., & Brooks, R. (2005). An analysis of single stock futures trading in the U.S. Financial Services Review, 14(2), 8595. Lee, C., & Tong, H. C. (1998). Stock futures: The effects of their trading on the underlying stocks in Australia. Journal of Multinational Financial Management, 8(23), 285301. Lehmann, B. N. (2002). Some desiderata for the measurement of price discovery across markets. Journal of Financial Markets, 5, 259276. Martens, M., Kofman, P., & Vorst, T. C. F. (1998). A threshold error-correction model for intraday futures and index returns. Journal of Applied Econometrics, 13, 245263. Manaster, S., & Rendleman, R. J., Jr. (1982). Option prices as predictors of equilibrium stock prices. The Journal of Finance, 37, 10431057. Tse, Y. (1999). Round-the-clock market efciency and home bias: Evidence from the international Japanese government bonds futures markets. Journal of Banking and Finance, 23, 18311860. Tse, Y. (2000). Further examination of price discovery on the NYSE and regional exchanges. Journal of Financial Research, 23, 331351. Tse, Y., & Erenburg, G. (2003). Competition for order ow, market quality, and price discovery in the Nasdaq 100 Index Tracking Stock. The Journal of Financial Research, 26, 301318.