The stock closing calland futures price behavior ... - Wiley Online Library

THE STOCK CLOSING CALL AND FUTURES PRICE BEHAVIOR: EVIDENCE FROM THE TAIWAN FUTURES MARKET HSIU-CHUAN LEE CHENG-YI CHIEN YEN-SHENG HUANG

This study examines the behavior of futures prices around stock market close before and after changes to the batching period of the stock closing call. On July 1, 2002, the Taiwan Stock Exchange expanded the length of the batching period roughly 10-fold, from an average of 30 seconds to 5 minutes. This change presents an opportunity to analyze how a stock closing method affects the behavior of index futures prices. Empirical results indicate that an increase in the length of the batching period affects the return volatility and trading volume of index futures contracts The authors would like to thank the Editor, Robert I. Webb, anonymous referees, and Angela Chen for their insightful comments and suggestions. Correspondence author, Department of Business Administration at National Taiwan University of Science and Technology in Taipei, Taiwan, ROC; e-mail: [email protected] Received September 2005; accepted November 2006

■

Hsiu-Chuan Lee is a PhD candidate in the Department of Business Administration at National Taiwan University of Science and Technology in Taipei, Taiwan, ROC.

■

Cheng-Yi Chien is a PhD candidate in the Department of Business Administration at National Taiwan University of Science and Technology in Taipei, Taiwan, ROC.

■

Yen-Sheng Huang is a Professor in the Department of Business Administration at National Taiwan University of Science and Technology in Taipei, Taiwan, ROC.

The Journal of Futures Markets, Vol. 27, No. 10, 1003–1019 (2007) © 2007 Wiley Periodicals, Inc. Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/fut.20284

1004

Closing Call

around stock market close. Furthermore, preclose stock returns have a great impact on extended futures returns when the batching period of the stock closing call is long. © 2007 Wiley Periodicals, Inc. Jrl Fut Mark 27:1003–1019, 2007

INTRODUCTION This study examines the impact of an increase in the length of the batching period of the stock closing call on futures price behavior. The contagion model suggests that futures traders make trades according to the information available for stock prices around the stock market close (see Chang, Jain, & Locke, 1995; Fong & Frino, 2001). Accordingly, it is likely that a substantial change in the length of the batching period of the stock closing call will affect the behavior of futures prices as the market closes.1 The issue of how the introduction of a stock closing call affects stock price behavior has received much attention in recent years. Pagano and Schwartz (2003) found that it reduces execution costs for individual traders and sharpens price discovery for the broad market. Hillion and Suominen (2004) suggested that introducing a closing call can reduce manipulation of the closing price. However, the price behavior of the futures market near stock market close is also importantly related to the intraday pattern of futures prices. Chang et al. (1995), Daigler (1997), Fong and Frino (2001), and Huang (2002) have all found that return volatility and trading volume for futures contracts decline after the stock market closes. Although previous research has studied futures price behavior near stock market close, no work has been done on the influence of a change in the length of the batching period of the stock closing call on futures price behavior. On July 1, 2002, the Taiwan Stock Exchange (TSE) expanded the length of the batching period of the stock closing call roughly 10-fold, from an average of 30 seconds to 5 minutes. This substantial change in the length of the batching period presents an opportunity to examine how the stock closing method affects the behavior of futures prices. Empirical results indicate that return volatility and trading volume for futures contracts decline significantly after the stock market closes when the batching period is short. In contrast, with a long batching period, 1

Amihud, Mendelson, and Lauterbach (1997) suggested that investors can easily use the observed prices from a primary security to infer the value of a secondary security during a continuous auction. Lang and Lee (1999) found that when the batching period is shortened, a call market method approaches a continuous market method. This suggests that futures traders can more easily make trades according to information for stock prices near stock market close in a short batching period than in a long one.

Journal of Futures Markets

DOI: 10.1002/fut

Lee, Chien, and Huang

return volatility and trading volume increase significantly after the market closes. Additionally, with the long batching period, preclose stock returns have a strong impact on futures returns in the extended futures trading period. The next section of this article presents a literature review, which is followed by a discussion of the institutional background of the stock market and futures market in Taiwan. Subsequent sections develop our hypotheses, describe the sample and methodology, and report empirical results. The last section presents our conclusion. LITERATURE REVIEW Numerous studies have examined the relationship between the stock market and futures market around stock market close. Chang et al. (1995) employed the contagion model to develop predictions for futures price behavior near stock market close. The contagion model, which was developed by King and Wadhwani (1990), states that trading in one market can change price behavior in other related markets because traders’ decisions will be influenced by observation of the primary market’s price behavior. Therefore, price movements in one market affect those in related markets. Chang et al. (1995) analyzed the impact of the closing of the New York Stock Exchange (NYSE) on Standard & Poor’s stock index futures. Using data from May 1982 through December 1990, they observed a decline in futures return volatility immediately after the NYSE closed. They suggested that this finding is consistent with the contagion effect, in that traders are using information from the spot market to judge the futures market near stock trading close. Daigler (1997) also utilized the contagion model and extended Brock and Kleidon’s (1992) market closure theory to address the intraday price behavior of futures markets.2 Daigler found that return volatility and trading volume tend to be high at the opening and closing periods for cash and futures markets. However, they decline for futures markets after the stock market closes. Fong and Frino (2001) investigated the impact of cash market closure on futures price behavior. Using data from the Hang Seng Index futures contracts, they found a decrease in futures return volatility after the stock market closes, and concluded that this is consistent with the contagion hypothesis. Huang (2002) examined intraday trading activity on the Taiwan Futures Exchange (TAIFEX) and in 2

Brock and Kleidon (1992) developed a market closure theory in which the large bid-ask spread, volatility, and volume that are frequently observed at both market open and close reflect the need of traders to rebalance portfolios after the market opens and before it closes. Journal of Futures Markets

DOI: 10.1002/fut

1005

1006

Closing Call

Singapore Exchange Derivatives Trading Limited. Huang indicated that return volatility and trading volume for the futures contracts traded on the TAIFEX decline after stock market close. Other research has examined the impact of the shift from a continuous close to a closing call on market quality and liquidity. Pagano and Schwartz (2003) examined the impact of the stock closing call on market quality for the Paris Bourse (now known as Euronext Paris). Utilizing data from the period 1995–1999, they found that introducing the closing call reduces execution costs for individual participants and sharpens price discovery for the broad market. Aitken, Comerton-Forde, and Frino (2005) studied the impact of a closing call auction on market liquidity. Using data from August 9, 1996 through August 10, 1997, they reported a significant decrease in the percentage of daily volume traded in the last hour following the introduction of the closing call auction. Additionally, they found that investors usually delay their trades until the closing call auction. In summary, previous research has investigated futures price behavior near stock market close and the impact of introducing a stock closing call on market quality and liquidity. However, the impact of a change in the length of the batching period of the stock closing call on futures price behavior has not been analyzed. Accordingly, this study attempts to fill the gap. INSTITUTIONAL BACKGROUND The TSE is an order-driven call market without designated market makers. Unlike the NYSE or the National Association of Securities Dealers Automated Quotation (NASDAQ) in the United States, it has no specialists or dealers involved in the process of trading. Public traders submit limit orders to brokers via the telephone, computer networks, or order forms filled out directly in the brokerage houses.3 Orders placed by public traders are sent by brokers to the fully automated securities trading system (FASTS) of the stock exchange for clearing. Thus, orders must be placed by brokers rather than directly by the stock exchange. The stock exchange accepts only limit orders, restricted to day orders that expire at the end of each trading day. Other orders, such as good-till-cancelled orders, are not available (see Chang, Hsu, Huang, & Rhee, 1999; Rhee & Wang, 1997). Traders can place both buy orders and sell orders, and they are free to cancel them. Generally, order placement or cancellation can 3

The TSE imposes a daily price limit of 7%. Market orders submitted by public traders are converted to highest buy limit orders or lowest sell limit orders by their brokers.


DOI: 10.1002/fut


be executed within seconds. Short selling is available, but is restricted to certain approved stocks. The TSE uses a periodical batch process to determine clearing prices throughout the whole trading period.4 For each round of market clearing, all buy orders with prices higher than the market clearing price and all sell orders with prices lower than the market clearing price must be filled. Additionally, all buy and sell orders with prices equal to the market clearing price must be filled. In the sample period 2001–2003, the TSE began trading at 9:00 A.M. and ended at 13:30 P.M. To determine the opening price, buy and sell orders were submitted and accumulated during the half-hour period before opening, from 8:30 A.M. to 9:00 A.M. Opening price was determined by selecting the price that maximized matched trading volume. For intraday trading, buy and sell orders were accumulated and matched on an average interval of around 30 seconds by the same call market method that maximized trading volume. For the closing price, however, the TSE used different call intervals to determine the clearing price at different times during the sample period. Prior to July 1, 2002, closing price was determined by maximizing matched orders accumulated during a time interval of about 30 seconds prior to close. After July 1, 2002, the interval was increased to 5 minutes. For the index futures market, the underlying assets of index futures contracts traded in the TAIFEX are based on the capitalization-weighted market index of the TSE. Futures contracts on the TAIFEX began trading on July 21, 1998. The contract value of the index futures on the TAIFEX is the capitalization-weighted market index of the TSE multiplied by NT$ 200 (New Taiwan Dollars).5 The TAIFEX opens 15 minutes earlier and closes 15 minutes later than the TSE. Thus, trading hours for the TAIFEX were from 8:45 A.M. to 13:45 P.M. in the sample period, 2001–2003. The trading mechanism of the TAIFEX resembles that of the TSE in numerous ways. The TAIFEX uses an electronic trading system to determine market clearing prices. No market makers, such as specialists or dealers, are involved. Traders submit buy and sell orders to brokers. Again, only limit orders are permitted and traders are free to place or cancel orders during trading. These orders are then sent by brokers to the electronic trading system of the futures exchange for clearing. 4

To enhance the transparency of the stock market, from January 1, 2003, the TSE has disclosed information about unfilled buy orders and sell orders for order prices up to the closest five ticks surrounding the clearing price in each round of batching. 5 Transaction costs involve a transaction tax of 0.025% on sell orders plus a 0.1% commission fee on both the buy and sell sides of trading. The daily price limit of 7% is imposed on index futures. Journal of Futures Markets

DOI: 10.1002/fut

1007

1008

Closing Call

As with the stock exchange, the TAIFEX uses a call auction method to determine clearing prices in each batching round. For opening futures prices, buy and sell orders are batched for 15 minutes from 8:30 A.M. to 8:45 A.M. Following opening, the batching period in the intraday trading session was around 10 seconds.6 The batching period for the closing price was 5 minutes, from 13:40 P.M. to 13:45 P.M., in the sample period. Additionally, the TAIFEX discloses unfilled orders up to five ticks surrounding the clearing price in each matching round.

HYPOTHESES Admati and Pf leiderer (1988) developed an asymmetric information model to explain the trading behavior of informed and liquidity traders and how their trading affects volatility and trading volume. This model assumes that liquidity traders attempt to minimize adverse selection costs, whereas informed traders want to time their trades to maximize advantage. Admati and Pfleiderer’s model proposes that all strategic traders will choose to trade during the same periods and that trading volume and volatility will be high in these periods. Thus, the model suggests that return volatility and trading volume increase as information asymmetry is reduced. The patterns of return volatility and trading volume for futures contracts around stock market close are likely to change when the batching period of the stock closing call is changed. With a short batching period of about 30 seconds, information regarding stock prices is revealed gradually preceding the close of the stock market. By contrast, with a long batching period of 5 minutes, price information is not revealed until the stock market closes. Because the contagion effect implies that futures traders will use information regarding stock prices to make trades near stock market close (Chang et al., 1995), for futures traders a short batching period of the stock closing call is related to high revelation of stock price information around stock market close, and a long batching period is related to low information revelation. As suggested by Admati and Pfleiderer (1988), information revelation is positively correlated with return volatility and trading volume. Hence, Hypothesis 1 states: Hypothesis 1: When the batching period of the stock closing call is short, the return volatility and trading volume of futures contracts will decrease after stock market close; but when the batching period is long, they will increase. 6

On July 29, 2002, the TAIFEX adopted a continuous trading method.


DOI: 10.1002/fut


Preclose stock returns will have a strong impact on extended futures returns when the batching period of the stock closing call is long. Daigler (1997), Fong and Frino (2001), and Huang (2002) suggested that the cash market dominates the corresponding futures markets around stock trading close, implying that preclose stock returns will affect futures returns at that point. Hence, with a short batching period information regarding stock prices is immediately incorporated into the futures market during the interval before the stock market closes. With a long batching period, however, this information is not incorporated into the futures market until the stock market closes. Thus, this study expects the following hypothesis to hold: Hypothesis 2: When the batching period of the stock closing call is short, preclose stock returns have little impact on extended futures returns; but when the batching period is long, they have a great impact. SAMPLE AND METHODOLOGY Sample This study utilizes data from the Taiwan Weighted Stock Index traded on the TSE and from Taiwan Index Futures contracts traded on the TAIFEX. The sample period covers 2 years from July 3, 2001 through June 18, 2003 (or a total of 480 trading days). The first subperiod, from July 3, 2001 through June 30, 2002 (or a total of 240 trading days), corresponds to the short stock closing call of 30 seconds. The second subperiod, from July 1, 2002 through June 18, 2003 (also 240 trading days), corresponds to the long stock closing call of 5 minutes. Intraday data for the Taiwan Weighted Stock Index and the Taiwan Index Futures contracts are obtained from the TSE and TAIFEX, respectively. For futures contracts, nearby futures contracts are used. Nearby futures contracts are rolled over to the next nearby contract 5 trading days before expiration to avoid the potential expiration effect. Methodology Volatility and Trading Volume To estimate the return volatility of futures contracts the return series are first calculated for each one-minute interval in the trading day based on transaction price data.7 The standard deviation (SD) of returns is estimated 7

The intraday futures return over a one-minute period is calculated as Ri,t ln(Pi,t/Pi1,t)

where Pi,t represents the intraday closing price on day t in interval i. Journal of Futures Markets

DOI: 10.1002/fut

1009

1010

Closing Call

for every 5-minute interval during the trading day.8 For comparison, the Garman-Klass volatility measure (GKV) is also estimated for every 5-minute interval as follows (see Garman & Klass, 1980): Var(G K) 12 [ln(High) ln(Low)]2 [2 ln(2) 1] [ln(Open) ln(Close)]2

(1)

where High, Low, Open, and Close are the highest, lowest, closing, and open prices, respectively in each 5-minute interval during the trading day. The GKV is usually considered a superior measure of volatility because it includes more information than the SD. To examine the impact of a change in the length of the batching period of the stock closing call, the standardized standard deviation (SSD) is utilized to control for the potentially distorting effect of different market conditions in the two subperiods (Chan, 2005; Chan, Chung, & Johnson, 1995). The SSD is calculated by subtracting the mean SD from the SD and dividing by the corresponding standard error as follows: SSD (SD MSD)/SESD, where the mean SD, MSD, and the standard error of SD, SESD, are estimated from the 5-minute series of SD for each trading day. These SSDs are then averaged across trading days for each subperiod of the sample. The standardized Garman-Klass volatility (SGKV) and standardized trading volume (STV) are calculated similarly to the SSD. To test the equality of volatility and trading volume for different trading periods, both t tests and Wilcoxon tests are used in this study. Because the t test and the Wilcoxon test give similar results, this study only reports Wilcoxon test results. The Impact of Preclose Stock Returns on Extended Futures Returns The following GARCH (1,1) model is used to assess the impact of preclose stock returns, SR1330,t, on extended futures returns, FRt,:9 FRt a1 a2SR1330,t a3SR1330,t DUMt a4FR1330,t et ht b1 b2 e2t 1 b3 ht 1 8

(2)

Chang et al. (1995), Daigler (1997), Fong and Frino (2001), and Huang (2002) all used the 5-minute interval to investigate futures price behavior near stock market close. 9 The relationship between futures returns and stock returns in the interval before stock close are also examined. We find that preclose stock returns (13:25 P.M.–13:30 P.M.) have a great impact on futures returns (13:25 P.M.–13:30 P.M.) when the batching period of the stock closing call is short. In contrast, preclose stock returns (13:25 P.M.–13:30 P.M.) have a small impact on futures returns (13:25 P.M.–13:30 P.M.) when the batching period of the stock closing interval is long. The empirical results are not reported here, but are available upon request. Journal of Futures Markets

DOI: 10.1002/fut


where FRt is the futures return in the postclose interval on day t; SR1330,t is the stock return in the preclose interval 13:25 P.M. to 13:30 P.M. on day t; FR1330,t is the futures return in the preclose interval 13:25 P.M. to 13:30 P.M. on day t; DUMt is 1 for the subperiod from July 1, 2002 through June 18, 2003 (the long batching period) and 0 otherwise (the short batching period); et is the residual on day t, et | t ~ N(0, ht); and ht is the conditional variance on day t. For futures and spot returns, the natural log of price relatives, ln(Pt /Pt1), is used. The preclose futures returns, FR1330,t, are included to control for potential influence on FRt. When stock returns in the preclose interval, SR1330,t, have a great impact on extended futures returns in the second subperiod, the coefficient a3 will be significantly positive in the regression model. EMPIRICAL RESULTS Volatility and Trading Volume

0.60 0.40 0.20 0.00

Short batching period Long batching period

-0.20 -0.40 -0.60

13:45

13:40

13:35

13:30

13:25

13:20

13:15

-1.00

13:05

-0.80 13:10

Standardized Standard Deviation (SSD)

Figures 1 and 2 present the standardized return volatility around stock market close for the two subperiods. For the first, with a short batching period, the SSD and SGKV move downward following stock market close. This finding is consistent with those obtained by Chang et al. (1995), Daigler (1997), Fong and Frino (2001), and Huang (2002). The standardized return volatility exhibits a different pattern for the second subperiod when the batching period of the stock closing call was extended to 5 minutes Figures 1 and 2 indicate that the SSD and SGKV increase following stock market close for the second subperiod with a long batching period.

Time FIGURE 1

Standardized standard deviation near stock market close for the two subperiods. Journal of Futures Markets

DOI: 10.1002/fut

1011

Closing Call

0.60 0.40 0.20 0.00


-0.20 -0.40

13:45

13:40

13:35

13:30

13:25

13:20

13:15

-0.80

13:10

-0.60 13:05

Standardized Garman-Klass Volatility (SGKV)

1012

Time FIGURE 2

Standardized Garman-Klass volatility near stock market close for the two subperiods.

Table I reports Wilcoxon rank sum test results for equality of the SSD before and after stock market close for the two subperiods. Panel A of Table I indicates that the SSD reaches 0.45 in the 5-minute interval before stock market close (13:25 P.M.–13:30 P.M.) during the first sub-eriod with a short batching period. The SSD drops to 0.26 in the 5-minute interval after stock close (13:30 P.M.–13:35 P.M.). Panel B of Table I indicates that the SSD is significantly different in the 5-minute intervals before and after stock close; and the value of the Wilcoxon test is 7.95. This pattern changes dramatically for the second subperiod, when the batching period of the stock closing call was extended to 5 minutes. The SSD is 0.40 for the 5-minute interval before stock close. By contrast, it reaches a peak of 0.06 in the 5-minute interval after stock close. The Wilcoxon test indicates that the increase in the SSD is significantly positive, at 4.23, between the 5-minute intervals before and after stock close. Table II reports Wilcoxon rank sum test results for equality of the SGKV before and after stock market close for the two subperiods. In Panel B of Table II, the Wilcoxon test indicates that the decline in the SGKV is significantly negative, at 7.31, between the 5-minute intervals before and after stock close when the batching period is short. By contrast, the test indicates that the SGKV is significantly positive, at 3.91, between the 5-minute intervals before and after stock close when the batching period is long. Figure 3 and Table III present the STV near stock market close for the two subperiods. The STV declines after stock market close when the batching period is short. This finding is consistent with those obtained by Journal of Futures Markets

DOI: 10.1002/fut


TABLE I

The Standardized Standard Deviation of Futures Returns for the Two Subperiods Sample period sample size

Before 240 trading days

After 240 trading days

Wilcoxon rank sum test statistics

Panel A (P.M.)

SSD

SSD

Before and after

(1) 13:20–13:25 (2) 13:25–13:30 (3) 13:30–13:35 (4) 13:35–13:40 (5) 13:40–13:45

0.48 0.45 0.26 0.68 0.76

0.10 0.40 0.06 0.44 0.58

7.07*** 9.42*** 2.80*** 3.91*** 2.69***

Panel B Hypothesis

Wilcoxon rank sum test statistics for different trading periods 0.59 7.95***

(2) and (1) (3) and (2)

3.97*** 4.23***

Note. This table presents the standardized standard deviation (SSD) of futures returns near stock market close for the two subperiods. “Before” is from July 3, 2001 through June 30, 2002, when the batching period was short (30 seconds). “After” is from July 1, 2002 through June 18, 2003, when the batching period was long (5 minutes). The SSD is the standardized standard deviation of futures returns. *Significance at 10%.**Significance at 5%. ***Significance at 1%.

TABLE II

The Standardized Garman-Klass Volatility for the Two Subperiods Sample period sample size


Panel A (P.M.)

SGKV

After 240 trading days SGKV

Wilcoxon rank sum test statistics Before and after

0.14 0.36 0.13 0.38 0.60

(1) 13:20–13:25 (2) 13:25–13:30 (3) 13:30–13:35 (4) 13:35–13:40 (5) 13:40–13:45

0.31 0.36 0.24 0.48 0.70

5.25*** 8.99*** 1.98** 3.39*** 3.52***

Panel B Hypothesis

Wilcoxon rank sum test statistics for different trading periods

(2) and (1) (3) and (2)

0.43 7.31***

4.81*** 3.91***

Note. This table presents the standardized Garman-Klass volatility (SGKV) of futures prices near stock market close for the two subperiods. “Before” is from July 3, 2001 through June 30, 2002, when the batching period was short (30 seconds). “After” is from July 1, 2002 though June 18, 2003, when the batching period was long (5 minutes). The SGKV is the standardized Garman-Klass volatility of futures prices. *Significance at 10%. **Significance at 5% . ***Significance at 1% . Journal of Futures Markets

DOI: 10.1002/fut

1013

Closing Call

1.20 1.00 0.80 0.60


0.40 0.20

13:45

13:40

13:35

13:30

13:25

13:20

13:15

-0.20

13:10

0.00 13:05

Standardized Trading Volume (STV)

1014

Time FIGURE 3

Standardized trading volume near stock market close for the two subperiods.

TABLE III

The Standardized Trading Volume for the Two Subperiods Sample period sample size


After 240 trading days

Wilcoxon rank sum test statistics

Panel A (P.M.)

STV

STV

Before and after

(1) 13:20–13:25 (2) 13:25–13:30 (3) 13:30–13:35 (4) 13:35–13:40 (5) 13:40–13:45

0.55 0.74 0.49 0.66 0.91

0.21 0.04 0.69 0.89 1.13

3.86*** 8.17*** 3.40*** 3.18*** 2.84***

Panel B Hypothesis (2) and (1) (3) and (2)

Wilcoxon Rank Sum Test Statistics for Different Trading Periods 2.37** 2.87***

2.26** 9.15***

Note. This table presents the standardized trading volume (STV) of futures contracts for the two subperiods near stock market close. “Before” is from July 3, 2001 through June 30, 2002, when the batching period was short 30 seconds). “After” is from July 1, 2002 through June 18, 2003, when the batching period was long (5 minutes). The STV is the standardized trading volume of futures contracts. *Significance at 10% . **Significance at 5% . ***Significance at 1% .

Daigler (1997) and Huang (2002). By contrast, the STV increases after stock market close when the batching period is long. Panel B of Table III reports that, for the first subperiod with a short batching period, the STV is significantly lower in the 5-minute interval after stock close than in the 5-minute interval before; and the value of the Wilcoxon test is 2.87. Conversely, for the second subperiod, the STV is significantly higher in the 5-minute interval after stock close than in the 5-minute interval before; and the value of the Wilcoxon test is 9.15. Journal of Futures Markets

DOI: 10.1002/fut


The results of SSD, SGKV, and STV are consistent with Hypothesis 1. With a short batching period, return volatility and trading volume for futures contracts decline after stock market close. By contrast, with a long batching period they increase. The results suggest that return volatility and trading volume for futures contracts are positively correlated with information revelation regarding stock prices near stock market close. Hence, these empirical findings are consistent with the information asymmetry hypothesis suggested by Admati and Pfleiderer (1988). The Impact of Preclose Stock Returns on Extended Futures Returns Table IV presents the relationship between preclose stock returns and extended futures returns. The regression coefficients for SR1330,t DUMt are 0.50, 0.45, and 0.35 for the three time intervals 13:30 P.M.–13:35 P.M., 13:30 P.M.–13:40 P.M., and 13:30 P.M.–13:45 P.M., respectively. Regression coefficients for SR1330,t DUMt are all significant at the 1% level. These empirical results indicate that preclose stock returns have a great impact on extended futures returns for the second subperiod with the long batching period. Table IV indicates that futures returns from before stock close can immediately incorporate information from the stock market before the market closes when the batching period is short. A long batching period delays information flow from the cash market to the futures market. The empirical results suggest that the cash market dominates the corresponding futures market near stock market close. These findings are consistent with Hypothesis 2. Results for Short Sample Periods To examine the robustness of Hypotheses 1 and 2, empirical results are tested for shorter subperiods. The whole sample period is divided into four 120-trading-day subperiods to examine the behavior of futures volatility and trading volume. So, the original first subperiod, with the short batching period, is divided into two further subperiods: Pre1 and Pre2. Similarly, the second subperiod, with the long batching period, is divided into two further subperiods: Post1 and Post2. Pre1 and Pre2 cover the half years from July 3, 2001 through December 27, 2001 and from December 28, 2001 through June 30, 2002, respectively; Post1 and Post2 cover the half years from July 1, 2002 through December 18, 2002 and from December 19, 2002 through June 18, 2003, respectively. Journal of Futures Markets

DOI: 10.1002/fut

1015

1016

Closing Call

TABLE IV

The Impact of Preclose Stock Returns on Extended Futures Returns FRt

FRt

FRt

Coefficient t 13:30–13:35 P.M. t 13:30–13:40 P.M. t 13:30–13:45 P.M.

Variable Constant SR1330,t SR1330,t DUMt FR1330,t Intercept e2t1 ht1 Adjusted R 2 Durbin–Watson Log-likelihood Sample size

a1 a2 a3 a4 b1 b2 b3

0.00* 0.24*** 0.50*** 0.26*** 0.00 0.15* 0.60*** 0.24 2.04 2455.59 480

0.00*** 0.26*** 0.45*** 0.17*** 0.00*** 0.15*** 0.60*** 0.12 2.04 2269.79 480

0.00* 0.19** 0.35*** 0.16*** 0.00 0.03 0.82*** 0.03 2.08 2138.75 480

Note. FRt a1 a2SR1330,t a3SR1330,t DUMt a4FR1330,t et ht b1 b2 e2t1 b3ht1 The sample period covers 2 years from July 3, 2001 through June 18, 2003. FRt is the futures return in the postclose interval on day t ; SR1330,t is the stock return in the preclose interval 13:25–13:30 P.M. on day t; FR1330,t is the futures return in the preclose interval 13:25–13:30 P.M. on day t; DUMt is 1 in the subperiod from July 1, 2002 through June 18, 2003 and 0 otherwise; et is the residual on day t, et | t ~ N(0, ht ); ht is the conditional variance on day t. *Significance at 10% . **Significance at 5%. ***Significance at 1% .

Table V presents the standardized futures volatility and trading volume for these four subperiods. Panel D of Table V reports that the results of the Wilcoxon test of standardized volatility and trading volume are all negative during subperiods Pre1 and Pre2. This suggests that volatility and trading volume decrease following stock market close during Pre1 and Pre2. By contrast, Panel D of Table V reports that the results for subperiods Post1 and Post2 are all positive. This indicates that volatility and trading volume increase after stock market close during Post1 and Post2. Additionally, when stock returns, SR1330,t, have a large impact on extended futures returns in the second half of the sample period (when the batching period is long), the regression coefficients for preclose stock returns, a4 and a5, are expected to be significantly positive. Table VI reports the regression results for three intervals: 13:30 P.M.–13:35 P.M., 13:30 P.M.–13:40 P.M., and 13:30 P.M.–13:45 P.M. The regression coefficients for a4 and a5 are all significantly positive. In summary, the empirical results for volatility, trading volume, and regression analysis for the short sample periods are consistent with Hypotheses 1 and 2.


DOI: 10.1002/fut


TABLE V

The Impact of the Batching Period of the Stock Closing Call on Trading Activity: Robustness Analysis Sample period sample size

Pre1 120 trading days

Pre2 120 trading days

Post1 120 trading days

Post2 120 trading days

Panel A: SSD (P.M.) (1) 13:25–13:30 (2) 13:30–13:35

0.50 0.15

0.41 0.37

0.43 0.20

0.37 0.08

0.36 0.17

0.37 0.30

0.32 0.24

0.40 0.02

0.83 0.51

0.66 0.47

0.10 0.64

0.02 0.74

Panel B: SGKV (P.M.) (3) 13:25–13:30 (4) 13:30–13:35

Panel C: STV (P.M.) (5) 13:25–13:30 (6) 13:30–13:35

Panel D: Hypothesis (2) and (1) (4) and (3) (6) and (5)

Wilcoxon rank sum test statistics for different trading periods 5.26*** 4.52*** 2.31**

6.01*** 5.82*** 1.67*

2.02** 1.71** 5.26***

4.08*** 3.92*** 7.71***

Note. This table presents the standardized return volatility and trading volume near stock market close for four subperiods. The first half of the sample period, with a short batching period, is divided into two subperiods, Pre1 and Pre2. The second half, with a long batching period, is divided into two subperiods, Post1 and Post2. “Pre1” is from July 3, 2001 through December 27, 2001 and “Pre2” is from December 28, 2001 through June 30, 2002; “Post1” is from July 1, 2002 through December 18, 2002 and “Post2” is from December 19, 2002 through June 18, 2003. The SSD is the standardized standard deviation of futures returns. The SGKV is the standardized Garman-Klass volatility of futures prices. The STV is the standardized trading volume of futures contracts. *Significance at 10% . **Significance at 5%. ***Significance at 1% .

CONCLUSION This study examines the impact of a substantial change in the length of the batching period of the stock closing call on index futures price behavior. Using data from the index futures contracts traded on the TAIFEX, it compares return volatility and trading volume for index futures when the batching period of the stock closing call is short (30 seconds) and when it is long (5 minutes). The empirical results indicate that return volatility and trading volume decline after stock market close when the batching period is short. By contrast, when the batching period is long, return volatility and trading volume increase after stock market close. These findings are consistent with the information asymmetry hypothesis proposed by Admati and Pf leiderer (1988). The results also indicate that preclose stock returns


DOI: 10.1002/fut

1017

1018

Closing Call

TABLE VI

The Impact of Preclose Stock Returns on Extended Futures Returns: Robustness Analysis FRt FRt FRt Coefficient t 13:30–13:35 P.M. t 13:30–13:40 P.M. t 13:30–13:45 P.M.

Variable Constant SR1330,t SR1330,t Pre1t SR1330,t Post1t SR1330,t Post2t FR1330,t Intercept 2 et1 ht1 Adjusted R 2 Durbin–Watson Log-likelihood Sample size

a1 a2 a3 a4 a5 a6 b1 b2 b3

0.00* 0.26*** 0.01 0.59*** 0.44*** 0.27*** 0.00 0.15* 0.60*** 0.24 2.01 2456.10 480

0.00*** 0.15*** 0.29*** 0.69*** 0.51*** 0.20*** 0.00*** 0.15*** 0.60*** 0.12 2.04 2269.65 480

0.00* 0.07 0.29* 0.44*** 0.48*** 0.17** 0.00 0.04 0.79*** 0.04 2.09 2140.08 480

FRt 1 2SR1330,t 3SR1330,t Pre1t 4SR1330,t Post1t

Note.

5SR1330,t Post2t 6FR1330,t et

ht 1 2e2t1 3ht1

The sample period covers 2 years from July 3, 2001 through June 18, 2003. FRt is the futures return in the postclose interval on day t; SR1330,t is the stock return in the preclose interval 13:25–13:30 P.M. on day t ; FR1330,t is the futures return in the preclose interval 13:25–13:30 P.M. on day t; Pre1t is 1 in the subperiod from July 3, 2001 through December 27, 2001 and 0 otherwise; Post1t is 1 in the subperiod from July 1, 2002 through December 18, 2002 and 0 otherwise; Post2t is 1 in the subperiod from December 19, 2002 through June 18, 2003 and 0 otherwise; et is the residual on day t, et | t ~ N(0, ht); ht is the conditional variance on day t. *Significance at 10%. **Significance at 5%. ***Significance at 1% .

have a great impact on extended futures returns when the batching period is long. This is consistent with the contagion model, which states that the stock market will dominate the corresponding index futures market around stock market close more than at other times of the trading day. BIBLIOGRAPHY Admati, A.R., & Pfleiderer, P. (1988). A theory of intraday patterns: Volume and price variability. Review of Financial Studies, 1, 3–40. Aitken, M., Comerton-Forde, C., & Frino, A. (2005). Closing call auctions and liquidity. Accounting and Finance, 45, 501–518. Amihud, Y., Mendelson, H., & Lauterbach, B. (1997). Market microstructure and securities values: Evidence from the Tel Aviv Stock Exchange. Journal of Financial Economics, 45, 365–390. Brock, W.A., & Kleidon, A.W. (1992). Periodic market closure and trading volume: A model of intraday bids and asks. Journal of Economic Dynamics and Control, 16, 451–489. Chan, K., Chung, Y.P., & Johnson, H. (1995). The intraday behavior of bid-ask spreads for NYSE stocks and CBOE options. Journal of Financial and Quantitative Analysis, 30, 329–346. Journal of Futures Markets

DOI: 10.1002/fut


Chan, Y.C. (2005). Who trades in the stock index futures market when the underlying cash market is not trading? Pacific-Basin Finance Journal, 13, 547–561. Chang, E.C., Jain, P.C., & Locke, P.R. (1995). Standard & Poor’s index futures volatility and price changes around the New York Stock Exchange close. Journal of Business, 68, 61–84. Chang, R.P., Hsu, S.T., Huang, N.K., & Rhee, S.G. (1999). The effects of trading methods on volatility and liquidity: Evidence from the Taiwan Stock Exchange. Journal of Business Finance & Accounting, 26(1) & (2), 137–170. Daigler, R.T. (1997). Intraday futures volatility and theories of market behavior. Journal of Futures Markets, 17, 45–74. Fong, K., & Frino, A. (2001). Stock market closure and intraday stock index futures market volatility: “Contagion,” bid-ask bias or both? Pacific-Basin Finance Journal, 9, 219–232. Garman, M., & Klass, M. (1980). On the estimation of security price volatilities from historical data. Journal of Business, 53, 67–78. Hillion, P., & Suominen, M. (2004). The manipulation of closing prices. Journal of Financial Markets, 7, 351–375. Huang, Y.C. (2002). Trading activity in stock index futures markets: The evidence of emerging markets. Journal of Futures Markets, 22, 983–1003. King, M.A., & Wadhwani, S. (1990). Transmission of volatility between stock markets. Reviews of Financial Studies, 3, 5–33. Lang, L.H.P., & Lee, Y.T. (1999). Performance of various transaction frequencies under call markets: The case of Taiwan. Pacific-Basin Finance Journal, 7, 23–39. Pagano, M.S., & Schwartz, R.A. (2003). A closing call’s impact on market quality at Euronext Paris. Journal of Financial Economics, 68, 439–484. Rhee, S.G., & Wang, C.J. (1997). The bid-ask bounce effect and the spread size effect: Evidence from the Taiwan stock market. Pacific-Basin Finance Journal, 5, 231–258.


DOI: 10.1002/fut

1019