Printer Friendly

An Analysis of Intraday Quoted Bid-Ask Spreads in Futures Markets: Evidence from the Sydney Futures Exchange.

Abstract:

Prior research documents an elevation in bid-ask spreads at the open and close of trading in futures markets. These findings directly contradict prior literature examining option and equities markets organised as competitive dealer markets, which also document a widening in spreads at the open, but provide evidence of a narrowing at the close. While prior futures market literature has relied on various estimators of bid-ask spreads, this is the first study to provide evidence on intraday quoted bid-ask spreads in futures markets. The evidence reported in this paper is consistent with prior equities and options market literature, and suggests that the findings in prior futures market research is driven by the spread estimators used. The primary determinants of bid-ask spreads (volume & volatility) are both elevated at the open and close of trading, which is similar to patterns documented in prior research. These findings are consistent with the predictions of inventory holding and adverse selection cost models of spreads.

Keywords:

MICROSTRUCTURE; BID-ASK SPREADS; SYDNEY FUTURES EXCHANGE.

1. Introduction

Brock and Kleidon (1992), McInish and Wood (1992), and Chan, Chung and Johnson (1995) analyse intraday patterns in quoted bid-ask spreads on the New York Stock Exchange (NYSE), which is organised as a specialist dealer market. They all document a U-shaped pattern in intraday quoted bid-ask spreads, which mirror patterns in intraday volume and volatility--two well known determinants of bid-ask spreads. These findings are consistent with specialists using their market power to extract economic rents from traders at the open and close, when the demand to trade is most inelastic. Chun, Christie and Schultz (1995) analyse intraday patterns in quoted bid-ask spreads on NASDAQ--a competitive dealer market. In contrast to previous research, they find that inside dealer spreads narrow throughout the trading day, especially near the close. They argue that unlike NYSE specialists, NASDAQ dealers have no special knowledge of order flow and little market power, and conclude that the reduction in bid-ask spreads on NASDAQ results from dealer inventory management and the desire of dealers to `go home flat'. Their findings are broadly consistent with Chun, Chung and Johnson (1995) who examine bid-ask spreads for options traded on the Chicago Board Options Exchange (CBOE), which is also a competitive dealer market. Locals in futures markets can also be characterised as `competitive dealers', and prior research has demonstrated that they also `go home flat' (Manaster & Mann 1996). Hence, the intraday behaviour of quoted spreads in futures markets is expected to be similar to NASDAQ and the CBOE.

In apparent contradiction to the equities and options market research discussed above, Ma, Peterson and Sears (1992) and Wang et al. (1994) both document U-shaped intraday patterns in bid-ask spreads on the Chicago Mercantile Exchange (CME). Further, Wang et al. (1994) find that after controlling for the determinants of spreads, there is no longer an intraday pattern. If taken at face value, this implies that intraday spreads in futures markets behave differently to other competitive dealer markets. The aim of this paper is to determine whether bid-ask spreads in futures markets behave differently to competitive equities and options markets, and attempt to reconcile the apparent contradiction in the literature.

Smith and Whaley (1994) identify that prior futures markets research has had to rely on various estimators of bid-ask spreads because quote information is not collected on US exchanges. One of the reasons for the conflicting findings in prior work could be that while the equities and options market research analyses actual quoted spreads, prior futures market research has been reduced to examining various estimates of spreads.(1) The Sydney Futures Exchange (SFE) is organised similarly to US futures markets. However, price reporters on the floor of the SFE collect quotes as well as trade information. This provides a unique opportunity to extend prior research on quoted spreads to futures markets. In this study, we analyse the intraday behaviour of quoted bid-ask spreads on the SFE in attempting to reconcile previous literature.

The remainder of this paper is organised as follows. Section 2 reviews the theoretical work developed to explain intraday patterns in bid-ask spreads, and a discussion of their relevance in a futures market setting. Section 3 discusses the data and method used to document and analyse the patterns in intraday quoted bid-ask spreads. Section 4 sets out the results, while the last section concludes and provides some suggestions for future research.

2. Theory

There are two important features of the market structure of the SFE that are likely to influence intraday patterns in spreads. First, SFE locals, similar to CME locals, CBOE dealers and NASDAQ dealers compete against each other for order flow in the provision of liquidity services (see Chan, Chung & Johnson 1995; Chan, Christie & Schultz 1995). Hence, these markets can generally be described as competitive dealer markets. Second, unlike overseas markets, floor trading on the SFE breaks for lunch as well as overnight (see the appendix). In the remainder of this section we review the explanations provided for intraday patterns in bid-ask spreads, and predict how these important features of the SFE can interact with trading to influence bid-ask spreads.

The theoretical literature on bid-ask spreads has developed three broad explanations for intraday patterns: market power, inventory holding costs and information asymmetry; each of which will be discussed in turn.

2.1 Market Power

Brock and Kleidon (1992) develop a model which demonstrates that monopoly market makers, in using their market power to extract rents from traders, will widen spreads during periods of high transaction demand. Their model recognises that since the optimal portfolio holdings of market agents are likely to change immediately before and after a trading break (e.g. overnight), that transaction demand is also likely to increase. This implies that spreads and trading activity in specialist dealer markets can be elevated in the intervals around trading breaks. On competitive exchanges, where dealers do not have market power, this is not expected to occur. Empirical tests of the market power hypothesis revolve around comparisons of spreads immediately before and after the overnight trading break on the NYSE (a specialist dealer market) and other competitive dealer markets, including the CBOE (Chan, Chung & Johnson 1995) and NASDAQ (Chan, Christie & Schultz 1995). These papers demonstrate that spreads appear to widen before trading breaks on the NYSE, but not in competitive dealer markets. This literature implies that spreads are not expected to widen immediately before and after a trading break on competitive futures exchanges such as the SFE or CME.

2.2 Inventory Holding Costs

Inventory holding cost models of the bid-ask spread can be traced back to Stoll (1978) for specialist dealer markets, and Ho and Stoll (1980; 1983) and Ho and Macris (1985) for competitive dealer markets. Inventory holding cost models imply that the bid-ask spread is the premium paid to dealers for taking a position in a security in making a market, which causes them to deviate from their preferred (optimum) portfolio. Hence the model relates bid-ask spreads to inventory holding costs.

Chan, Chung and Johnson (1995) recognise that, in a specialist dealer market, these models imply a positive relationship between trading activity and bid-ask spreads as specialists may be required to hold larger inventory positions during periods of intense trading activity. Hence, bid-ask spreads may widen at the open and close of trading, reflecting the elevation in trading activity. However, in a competitive dealer market the reverse can occur. Since competitive dealers will find it easier to balance their inventory during periods of higher trading activity, spreads may narrow. This prediction is consistent with the theoretical work of Ho and Macris (1985). The empirical tests carried out by Chan, Chung and Johnson (1995) and Chan, Christie and Schultz (1995) also support this theoretical work in documenting a narrowing in bid-ask spreads before market closure for the CBOE and NASDAQ and a widening in spreads on the NYSE. This literature implies that bid-ask spreads on the SFE and CME should narrow immediately before a trading break.

2.3 Information Asymmetry

Following Bagehot (1971), a number of papers suggest that bid-ask spreads are set by market makers to offset expected losses from trading with informed traders against expected gains from trading with liquidity traders. Copeland and Galai (1983) were the first to provide a rigorous model of spreads based on information asymmetry considerations. Foster and Viswanathan (1994) develop an asymmetric information model which predicts elevated volume, volatility and bid-ask spreads at the start of the trading day. Madhavan (1992) also develops an information asymmetry model which implies that trading resolves uncertainty. Together these papers imply that bid-ask spreads are expected to be wide at the opening of trading and gradually decline throughout the trading day. As argued by Chan, Chung and Johnson (1995) and Chan, Christie and Schultz (1995) this is likely to be the main explanation for the broad intraday pattern that they document for the CBOE and NASDAQ, both competitive dealer markets. Consistent with this literature, it is anticipated that spreads will be elevated when trading resumes immediately following a trading break on the SFE and CME.

2.4 Empirical Implications

While the literature reviewed above generates a number of different predictions for bid-ask spreads in competitive dealer markets such as the CME and SFE when the market resumes for trading following a break, the predictions for the close of trading are similar. For the opening of trading, the information asymmetry model predicts an elevation in bid-ask spreads, while the inventory control model predicts a narrowing. For the close of trading, none of the models predict a widening in spreads. The prediction for the close of trading is directly contradicted by the findings of Ma, Peterson and Sears (1992) and Wang et al. (1994), which demonstrate that different estimates of spreads for futures traded on the CME are elevated immediately prior to the close of trading, while patterns in their determinants are similar to equities and option exchanges organised as competitive dealer markets. This suggests that spreads in futures markets behave differently to equities and options markets. In this paper we reexamine these hypotheses using actual quoted spreads for the SFE.

3. Data and Method

The data available for this study was provided by the SFE. The data describes trading in the SFE's four major futures contracts; the Share Price Index (SPI), 3-year bond contract, 10-year bond contract and 90-day bank accepted bill (BAB) contract. The data extends from 1 January 1992 to 31 December 1996 and includes trade and quote records which are time stamped to the nearest second. The trade records include fields describing the price and volume associated with trades, while the quote records document the price of bid and ask quotes. The procedure used by the SFE to collect this data is described in the appendix.

The size of the SPI and 90-day BAB contracts was changed on the 10 October 1993 and 1 May 1995 respectively, and a lunchtime trading experiment occurred for the SPI contract between 9 May 1994 and 30 September 1994.(2) The periods before and after the changes in the SPI and BAB contracts were analysed separately and the lunchtime trading period was excluded for the SPI. All analysis is carried out using a 10-minute observation interval.(3)

A time weighting procedure identical to McInish and Wood (1992) is employed to measure bid-ask spreads (SPREAD). Specifically, spreads were calculated as follows:

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where: [BAS.sub.i] = the quoted bid-ask spread in index or basis points;

[t.sub.i] = the amount of time the spread i exists; and

n = the number of different bid-ask spreads that occur during interval t.

Following Chan, Chung and Johnson (1995), the bid-ask spread is standardised by subtracting the mean and dividing by the standard deviation for the day on which spread t is observed.

To determine whether bid-ask spreads exhibit intraday patterns, we use a regression approach similar to prior research. The following model was estimated:

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where: ST([SPREAD.sub.t]) = the standardised bid-ask spread; and

[D.sub.k] = a time-of-day dummy variable which is set to one if the time-of-day for observation t corresponds to the k-th 10-minute interval of the day, otherwise zero.

The six time intervals from 2.30 p.m. to 3.20 p.m. were excluded for the purposes of constructing the time-of-day dummy variables, which implies that analysis of time-of-day patterns focuses on differences between these excluded intervals and surrounding intervals. The trading hours of the different contracts implies that there are 23 time-of-day dummy variables for the SPI contract and 33 for interest rate contracts.

The parameters of model 2 were estimated using the Generalised Method of Moments (GMM) technique pioneered by Hansen (1982), and applied in prior related research (see Foster & Viswanathan 1993; Chan, Chung and Johnson 1995; Chan, Christie and Schultz 1995). GMM is a robust estimator that does not require information on the exact distribution of the disturbances. The parameters of the model, which are given by the vector [Phi], need to satisfy a theoretical relation between some function of them, say f([Phi]), and a set of instrumental variables, Z. This relation, known as the orthogonality conditions, is given by E[f([Phi])' Z)] = 0. The GMM estimator selects parameter estimates so that the sample counterpart of the above orthogonality conditions hold; that is, so that the sample correlations between the function, f, and the instruments Z, are close to zero.(4)

One possible reason for the difference between spreads on the CME and SFE is a difference in the determinants of spreads. In order to eliminate differences in the determinants of spreads as a source for the differences in the behaviour of spreads in this paper and prior research, the time-of-day patterns in volume and volatility are documented. VOLATILITY is measured as the time-weighted standard deviation of bid-ask midpoints as follows (see McInish & Wood 1992):

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where: [t.sub.i] = the amount of time over which price Pi (measured using the midpoint of the spread) occurs over the observation interval;

n = the number of different prices that occur over the interval; and

P = the time-weighted average price.

Bid-ask midpoints are employed to avoid problems associated with bid-ask bounce, which upwardly biases volatility measured on the basis of transaction prices (Venkatesh 1992). Trading activity is proxied using contract volume (VOLUME) in each 10-minute interval.(5) All of these variables are then standardised by subtracting the mean and dividing by the standard deviation of the variable for the day on which observation t occurs. The GMM approach outlined above is also used to test the significance of intraday patterns in volume and volatility.

4. Results

4.1 Intraday Patterns in Bid-Ask Spreads

Figure 1 and table 1 illustrate the intraday pattern in quoted bid-ask spreads for each of the main futures contracts traded on the SFE. Figure 1 provides evidence of an elevation in bid-ask spreads at the commencement of trading across all contracts, except for the SPI contract in the second sample period. Table 1 confirms that these patterns are statistically significant. Consistent with the discussion in the theory section, there is no evidence of a widening in spreads immediately before the overnight and lunchtime trading break. In fact, figure 1 provides some evidence that spreads narrow in the intervals immediately before the overnight trading break. Table 1 again confirms that this pattern is statistically significant for all contracts except the BAB contract.

[Figure 1 ILLUSTRATION OMITTED]

Table 1 Generalised Method of Moments (GMM) Estimates of the Intraday Variation in Standardised Bid-Ask Spreads for Major SFE Futures Contracts(1)
 SPI

 1 Jan. 92 to 11 Oct. 93 to
 8 Oct. 93 31 Dec. 96

 Co-eff. t-stat. Co-eff. t-stat.

0-10 min. aftermarket
 Open 1.4326 19.31(*) -0.1897 -1.86
10-20 min. after Market
 Open 0.4222 6.99(*) 0.0259 0.55
20-30 min. after
 Market Open 0.2211 3.62(*) 0.0446 0.93
20-30 min. before
 Lunchtime Close -0.0831 -1.25 0.0902 1.74
10-20 min. before
Lunchtime Close -0.0303 0.43 0.1055 1.94
0-10 min. before Lunchtime
 Close -0.2179 -3.21(*) -0.0691 -1.29
0-10 min. after Lunchtime
 Open 0.7776 10.64(*) 0.3507 6.50(*)
10-20 min. after Lunchtime
 Open 0.2165 3.25(*) 0.1025 2.07
20-30 min. after Lunchtime
 Open 0.0735 1.19 0.0457 0.95
20-30 min. before
 Market Close -0.1356 -2.00 -0.1494 -2.93
10-20 min. before
 Market Close -0.3061 -4.81(*) -0.2944 -6.08(*)
0-10 min. before
 Market Close -0.5792 -9.37(*) -0.5920 -12.84(*)

Observations 11,442 22,533

 3-Year Bonds 10-Year Bonds

 Co-eff. t-stat. Co-eff. t-stat.

0-10 min. aftermarket
 Open 0.3681 6.71(*) 0.4195 7.42(*)
10-20 min. after Market
 Open 0.0731 1.70 0.0045 0.11
20-30 min. after
 Market Open 0.0494 1.01 0.0458 1.16
20-30 min. before
 Lunchtime Close 0.0394 0.95 -0.0146 -0.37
10-20 min. before
Lunchtime Close 0.0161 0.38 0.0258 0.35
0-10 min. before Lunchtime
 Close 0.0010 0.01 -0.1519 -3.65(*)
0-10 min. after Lunchtime
 Open 0.4526 9.70(*) 0.7442 15.31(*)
10-20 min. after Lunchtime
 Open 0.0873 2.08 0.1665 3.88(*)
20-30 min. after Lunchtime
 Open -0.0201 -0.47 0.1508 3.67(*)
20-30 min. before
 Market Close -0.0424 -1.03 -0.0404 -1.06
10-20 min. before
 Market Close -0.0557 -1.27 -0.0785 -2.07
0-10 min. before
 Market Close -0.1632 -3.65(*) -0.2488 -6.90(*)

Observations 49,062 49,062

 90-Day BAB

 1 Jan. 92 to 1 May 95 to
 31 Apr. 95 31 Dec. 96

 Co-eff. t-stat. Co-eff. t-stat.


0-10 min. aftermarket
 Open 1.9837 22.91(*) 1.0601 8.92(*)
10-20 min. after Market
 Open 0.3417 6.16(*) 0.1003 1.21
20-30 min. after
 Market Open 0.1260 2.39 0.1800 2.25
20-30 min. before
 Lunchtime Close 0.0994 1.42 0.1737 1.96
10-20 min. before
Lunchtime Close 0.0821 1.15 0.2032 1.70
0-10 min. before Lunchtime
 Close 0.1146 1.59 0.1627 1.55
0-10 min. after Lunchtime
 Open 0.7028 9.05(*) 0.5523 6.02(*)
10-20 min. after Lunchtime
 Open 0.3111 4.23(*) 0.2605 3.07
20-30 min. after Lunchtime
 Open 0.2435 3.34(*) 0.2154 2.59
20-30 min. before
 Market Close 0.1845 2.52 0.2879 3.14
10-20 min. before
 Market Close 0.1053 1.46 0.2470 2.37
0-10 min. before
 Market Close 0.0100 0.14 0.1288 1.50

Observations 26,364 16,419


Notes: (1.) This table reports selected coefficient estimates of the following model:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where ST([SPREAD.sub.t]) is the standardised time weighted quoted bid-ask spread (in ticks) occurring in interval t, and [D.sub.k] is a time-of-day dummy variable set to one if observation t falls in interval k, otherwise zero. The intervals spanning 2.30 p.m. to 3.20 p.m. are excluded in constructing time-of-day dummy variables. The significance level of each 10-minute interval is established using a t-statistic derived from GMM estimation of the model above. The t-statistics are adjusted for heteroskedasticity and autocorrelation.

(*) Significant at the 0.001 level.

Figure 1 also provides some evidence of a narrowing in spreads immediately before the lunchtime break, although table 1 suggests that they are only significantly lower for the first sample period for the SPI and 10-year bond futures. There is also clear evidence of a significant widening in spreads immediately following the lunchtime break for all contracts depicted in figure 1 and table 1. Hence, the behaviour in spreads around the lunchtime break is generally consistent with the overnight break.

The results presented so far are entirely consistent with the findings of Chan, Chung and Johnson (1995) and Chan, Christie and Schultz (1995) for quoted spreads in competitive equities markets. However, they directly contradict the findings of Wang et al. (1994) and Ma, Peterson and Sears (1992), who analysed various estimates of spreads for the CME and found evidence of a widening in spreads immediately before a trading break.(6) While this difference in results may stem from the nature of spreads analysed (i.e. quoted v. estimated spreads), it may also result from differences in intraday patterns in the determinants of spreads. These are discussed in the next section.

Another noteworthy feature of the results reported above is the increase in bid-ask spreads for the 10-minute interval ending at 11.40 a.m. Although not reported, the result is statistically significant at the 0.001 level. This corresponds to the time-of-day when most macroeconomic information is released in Australia (11.30 a.m.). The widening in bid-ask spreads around the announcement time is consistent with prior research analysing quoted spreads around information releases in equity markets (see Jennings 1994).

4.2 Intraday Patterns in the Determinants of Bid-Ask Spreads

4.2.1 Volume Both the equities and futures market literature cited earlier has documented that volume is elevated immediately prior to and following the overnight trading break (see Ma, Peterson & Sears 1992; Wang et al. 1994; Chan, Chung & Johnson 1995; Chan, Christie & Schultz 1995). Figure 2 and table 2 document the intraday pattern in trading volume for all major futures contracts traded on the SFE.

[Figure 2 ILLUSTRATION OMITTED]

Table 2 Generalised Method of Moments (GMM) Estimates of the Intraday Variation in Standardised Trading Volume for Major SFE Futures Contracts(1)
 SPI

 1 Jan. 92 to 11 Oct. 93 to
 8 Oct. 93 31 Dec. 96

 Co-eff. t-stat. Co-eff. t-stat.

0-10 min. after
 Market Open 1.9259 30.75(*) 2.2321 54.54(*)
10-20 min.
 after Market Open 1.4420 25.45(*) 1.3319 34.00(*)
20-30 min. aftermarket
 Open 1.1206 19.94(*) 1.1580 29.00(*)
20-30 min. before
 Lunchtime Close -0.1142 -2.13 -0.1817 -4.69(*)
10-20 min. before
 Lunchtime Close -0.2033 -3.90(*) -0.3651 -9.89(*)
0-10 min. before
 Lunchtime Close -0.1835 -3.55(*) -0.4095 -11.50(*)
0-10 min. after
 Lunchtime Open -0.0177 -0.31 0.3726 8.00(*)
10-20 min. after
 Lunchtime Open -0.1379 -2.71 -0.1137 -3.15
20-30 min. after
 Lunchtime Open -0.0304 -0.64 -0.0670 -2.11
20-30 min. before
 Market Close 0.0222 0.42 -0.0522 -1.39
10-20 min. before
 Market Close 0.3381 6.21(*) 0.1740 4.58(*)
0-10 min. before
 Market Close 1.7104 30.27(*) 1.3214 33.48(*)
Observations 11,442 22,533

 3-Year Bonds 10-Year Bonds

 Co-eff. t-stat. Co-eff. t-stat.

0-10 min. after
 Market Open 2.0404 47.41(*) 2.4435 60.30(*)
10-20 min.
 after Market Open 1.0020 26.48(*) 1.1979 33.76(*)
20-30 min. aftermarket
 Open 0.7409 19.64(*) 0.8603 24.78(*)
20-30 min. before
 Lunchtime Close 0.1636 4.47(*) 0.2507 6.87(*)
10-20 min. before
 Lunchtime Close 0.0514 1.47 0.1049 2.94
0-10 min. before
 Lunchtime Close 0.1354 3.91(*) 0.1580 4.59(*)
0-10 min. after
 Lunchtime Open 0.2206 5.59(*) 0.2361 6.01(*)
10-20 min. after
 Lunchtime Open 0.0332 0.94 0.0826 2.34
20-30 min. after
 Lunchtime Open 0.0785 2.24 0.1161 3.40(*)
20-30 min. before
 Market Close 0.3758 10.41(*) 0.4030 11.74
10-20 min. before
 Market Close 0.4640 12.87(*) 0.5609 15.75
0-10 min. before
 Market Close 1.1446 29.14(*) 1.3099 33.06(*)
Observations 49,062 49,062

 90-Day BAB

 1 Jan. 92 to 1 May 95 to
 31 Apr. 95 31 Dec. 96

 Co-eff. t-stat. Co-eff. t-stat.

0-10 min. after
 Market Open 1.1991 16.70(*) 1.6641 18.52(*)
10-20 min.
 after Market Open 0.5998 10.06(*) 0.5602 8.04(*)
20-30 min. aftermarket
 Open 0.4978 8.81(*) 0.4610 6.60(*)
20-30 min. before
 Lunchtime Close -0.0702 -1.37 0.0140 0.23
10-20 min. before
 Lunchtime Close -0.0987 -1.98 -0.0853 -1.51
0-10 min. before
 Lunchtime Close -0.0140 -0.27 0.1087 1.75
0-10 min. after
 Lunchtime Open -0.1591 -3.15 0.0214 0.33
10-20 min. after
 Lunchtime Open -0.1256 -2.49 -0.0535 -0.84
20-30 min. after
 Lunchtime Open -0.1730 -3.64(*) -0.0918 -1.70
20-30 min. before
 Market Close 0.0945 1.78 0.0066 0.11
10-20 min. before
 Market Close 0.1501 2.85 0.1848 3.04
0-10 min. before
 Market Close 0.8410 13.64(*) 0.8835 11.99(*)
Observations 26,364 16,419


Notes: (1.) This table reports selected coefficient estimates of the following model:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where ST([VOLUME.sub.t]) is the standardised volume (in contracts) traded during interval t, and [D.sub.k] is a time-of-day dummy variable set to one if observation t falls in interval k, otherwise zero. The intervals spanning 2.30 p.m. to 3.20 p.m. are excluded in constructing time-of-day dummy variables. The significance level of each 10-minute interval is established using a t-statistic derived from GMM estimation of the model above. The t-statistics are adjusted for heteroskedasticity and autocorrelation.

(*) Significant at the 0.001 level.

Consistent with prior literature, figure 2 suggests that trading volume is elevated at the commencement of trading and at the end of the trading day across all contracts. The GMM results reported in table 2 suggest that the elevation in volume is statistically significant at the 0.001 level across all contracts. In summary, these findings are entirely consistent with the patterns documented in both the equities and futures market literature and, hence, cannot explain the difference in findings relating to spreads in this paper and prior futures market literature.

The behaviour of trading volume around the lunchtime interval is, however, quite different. The evidence in relation to trading volume immediately before and after the lunchtime period is at best mixed. Table 2 and figure 2 provide some evidence of a statistically significant increase in trading volume for 3-year bonds and 10-year bonds immediately prior to the lunchtime break. However, figure 2 suggests that volume is at its lowest intraday level for the SPI contract in both sample periods, while table 2 confirms that this is statistically significant at the 0.001 level. Similarly, following the lunchtime break, figure 2 suggests that volume is significantly higher for the SPI contract in the earlier sample period, however, no clear pattern emerges for the other futures series. Table 2 implies that trading volume is significantly higher for the SPI contract in the second sample period, as well as the 3-year bond and 10-year bond contract immediately following the lunchtime break. These results imply that, while the behaviour in bid-ask spreads around the lunchtime trading interval may be consistent with the overnight trading break, the forces shaping these patterns are significantly different.

4.2.2 Volatility The patterns documented in volatility in prior literature tend to be consistent with volume. That is, volatility is elevated immediately prior to and following the overnight trading break (see Ma, Peterson & Sears 1992; Wang et al. 1994; Chan, Chung & Johnson 1995; Chan, Christie & Schultz 1995). Figure 3 and table 3 document the intraday pattern in price volatility for all major futures contracts traded on the SFE. Figure 3 clearly suggests that price volatility is elevated around the overnight trading break. Table 3 confirms that these patterns are statistically significant at the 0.01 level across all contracts. The findings are entirely consistent with all prior literature, and again suggest that differences in price volatility cannot account for the conflicting findings between this study and prior futures market research.

[Figure 3 ILLUSTRATION OMITTED]

Table 3 Generalised Method of Moments (GMM) Estimates of the Intraday Variation in Standardised Price Volatility for Major SFE Futures Contracts(1)
 SPI

 1 Jan. 92 to 11 Oct. 93 to
 8 Oct. 93 31 Dec. 96
 Co-eff. t-stat Co-eff. t-stat

 0-10 min. after Market
 Open 1.3788 19.02(*) 1.1211 21.16(*)
10-20 min. after Market
 Open 0.9883 13.78(*) 0.8512 16.55(*)
20-30 min. after Market
 Open 0.7410 10.46(*) 0.6685 13.53(*)
20-30 min. before
 Lunchtime Close -0.1411 -2.25 -0.0517 -1.12
10-20 min. before
 Lunchtime Close -0.1985 -3.14 -0.2250 -5.07(*)
 0-10 min. before
 Lunchtime Close -0.2983 -4.80(*) -0.3902 -8.76(*)
 0-10 min. after
 Lunchtime Open 0.1147 1.77 0.1579 3.18
10-20 min. after
 Lunchtime Open -0.0529 -0.83 -0.0369 -0.81
20-30 min. after
 Lunchtime Open -0.0902 -1.49 -0.0245 -0.55
20-30 min. before Market
 Close -0.1522 -2.52 -0.0880 -1.93
10-20 min. before Market
 Close 0.0905 1.41 0.0029 0.06
 0-10 min. before Market
 Close 0.3397 5.22(*) 0.1546 3.22(*)
Observations 11,442 22,533

 3-Year Bonds 10-Year Bonds

 Co-eff. t-stat Co-eff. t-stat
 0-10 min. after Market
 Open 1.5297 35.17(*) 1.5246 34.11(*)
10-20 min. after Market
 Open 0.6953 20.38(*) 0.6953 18.10(*)
20-30 min. after Market
 Open 0.5342 14.52(*) 0.5342 13.64(*)
20-30 min. before
 Lunchtime Close 0.1094 4.23(*) 0.1094 2.89
10-20 min. before
 Lunchtime Close 0.2005 1.03 0.2005 1.33
 0-10 min. before
 Lunchtime Close -0.0313 1.11 -0.0313 -0.88
 0-10 min. after
 Lunchtime Open 0.3439 10.62(*) 0.3439 8.68(*)
10-20 min. after
 Lunchtime Open 0.0502 5.11(*) 0.0502 1.37
20-30 min. after
 Lunchtime Open 0.0806 4.44(*) 0.0806 2.23
20-30 min. before Market
 Close 0.3051 7.84(*) 0.3051 8.28(*)
10-20 min. before Market
 Close 0.3400 7.68(*) 0.3400 8.90(*)
 0-10 min. before Market
 Close 0.5296 10.51(*) 0.5296 13.18(*)
Observations 49,062 49,062

 90-Day BAB

 1 Jan. 92 to 1 May 95 to
 31 Apr. 95 31 Dec. 96

 Co-eff. t-stat Co-eff. t-stat
 0-10 min. after Market
 Open 1.8597 27.80(*) 1.7255 19.97(*)
10-20 min. after Market
 Open 0.6981 12.40(*) 0.5599 8.18(*)
20-30 min. after Market
 Open 0.4255 8.36(*) 0.3754 5.93(*)
20-30 min. before
 Lunchtime Close 0.0420 0.81 0.0952 1.71
10-20 min. before
 Lunchtime Close 0.0056 0.12 0.0765 1.29
 0-10 min. before
 Lunchtime Close -0.0324 -0.66 0.0615 1.12
 0-10 min. after
 Lunchtime Open 0.3090 5.92(*) 0.3832 5.80(*)
10-20 min. after
 Lunchtime Open 0.1222 2.47(*) 0.1563 2.60
20-30 min. after
 Lunchtime Open 0.0641 1.35 0.1122 1.88
20-30 min. before Market
 Close 0.1649 3.31(*) 0.1095 1.81
10-20 min. before Market
 Close 0.1833 3.62(*) 0.1712 2.78
 0-10 min. before Market
 Close 0.4582 8.69(*) 0.4445 6.55(*)
Observations 26,364 16,419


Notes: (1.) This table reports selected coefficient estimates of the following model:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where ST([VOLATILITY.sub.t]) is the standardised time weighted standard deviation of the quote midpoint for interval t, and Dk is a time-of-day dummy variable set to one if observation t falls in interval k, otherwise zero. The intervals spanning 2.30 p.m. to 3.20 p.m. are excluded in constructing time-of-day dummy variables. The significance level of each 10-minute interval is established using a t-statistic derived from GMM estimation of the model above. The t-statistics are adjusted for heteroskedasticity and autocorrelation.

(*) Significant at the 0.001 level.

Again, the behaviour in price volatility around the lunchtime break is substantially different to its behaviour around the overnight trading break. In contrast to the intervals immediately prior to the overnight trading break, figure 3 suggests that price volatility is at its intraday lowest for the SPI contract, and is close to being the lowest for the interest rate futures contracts. Table 3 reports that, in contrast to interest rate futures, price volatility for SPI futures are significantly lower immediately prior to the lunchtime break. While figure 3 appears to suggest that price volatility is elevated when trading resumes following the lunchtime break, table 3 reports that this is statistically significant for the interest rate futures contracts only.

5. Conclusions and Suggestions for Future Research

This paper documents an elevation in quoted bid-ask spreads at the opening of trading and some evidence of a narrowing in spreads immediately prior to the close. All of these findings are entirely consistent with prior equities and options market research examining quoted bid-ask spreads, and contrasts with futures market research examining various spread estimators. Further, two main determinants of bid-ask spreads, trading volume and volatility, are elevated at the open and close of trading. This is entirely consistent with all prior research. Hence, there is little evidence to suggest that spreads in futures markets behave differently to spreads in equities markets, and the findings in prior futures market research appear to be attributable to the estimators used. We concur with the conclusions of prior equities and options market research, that spreads at the open and close of trading in competitive securities markets are consistent with information asymmetry and inventory holding cost models.

While the behaviour of bid-ask spreads around the lunchtime closure are consistent with their behaviour around the overnight closure, patterns in volume and volatility are significantly different. This implies that the forces shaping spreads around the lunchtime closure are substantially different to the overnight closure. This finding is likely to be driven by the differences between the overnight and lunchtime trading break. The period of closure is brief (relative to the overnight close), volume and volatility in the underlying markets are expected to be at their lowest (Aitken, Brown & Walter 1994), it is unlikely that there will be official information releases during the period of closure, and the underlying markets remain open.

A number of possible future research directions are suggested by the findings in this paper. The results reported in this paper surrounding the lunchtime trading break contrast sharply with the results from US markets cited earlier which operate without a trading break. This implies that lunchtime closure can have an impact on market quality. A more detailed analysis of the lunchtime break can lead to insights relative to desirability of lunchtime closure. Such research is particularly important given the diversity in practice around the world. The SFE is not the only major exchange in the world with a lunchtime closure. The Tokyo International Financial Futures Exchange, the Osaka Securities Exchange and the Tokyo Stock Exchange also close at lunchtime, which is in contrast to the NYSE, the Chicago Board of Trade and the Chicago Mercantile Exchange which trade continuously.

Another possible future research direction is an analysis of bid-ask spreads and trading volume around information announcements. While a number of prior papers have examined the behaviour of intraday price volatility around information announcements in analysing the speed of response of futures markets to new information (e.g. Ederington & Lee 1993), data limitations have resulted in an absence of prior research examining bid-ask spreads and trading volume in futures markets. The findings in this paper suggest that bid-ask spreads and trading volume seem to react, and more detailed analysis can lead to further insight.

(1.) Another possible reason for the contrasting results are the informational differences between the equities/options markets and futures markets. It is well documented in prior literature that futures markets lead equities and options markets in terms of price discovery (e.g. Chun 1992; Fleming, Ostdiek & Whaley 1996). Such differences could manifest themselves in spreads, and generate the observed widening in spreads at the end of the day documented in prior futures market research. We thank one of the anonymous referees for suggesting this.

(2.) The size of the SPI contract was reduced from $100 multiplied by the index to $25 multiplied by the index on the 10 October 1993 while the value of the Bank Accepted Bill contract was doubled from $500,000 to $1,000,000 on 1 May 1995.

(3.) Ten minute periods are chosen, since trading of the SPI contract commences (and concludes) approximately 10 minutes before opening (and after closure) of the ASX.

(4.) The resultant criterion function which is minimised is defined as J = m([Phi])'Am([Phi]), where m([Phi]) = f([Phi])'Z, A is a weighting matrix which is symmetric positive definite, and m is the sample moments. A necessary condition to obtain (asymptotically) efficient estimates of [Phi] is that A is the inverse of the covariance matrix of the sample moments, m. The Bartlett kernel ensures the covariance matrix of the sample moments is positive semi-definite while the fixed bandwidth is selected using the Newey and West (1994) nonparametric method which adjusts for serial correlation.

(5.) The variable was also broken down into average trade size and trading frequency, and the time series for each of these examined separately. While the intraday patterns in both of these variables were consistent with the volume variable reported in the paper, the patterns in trade size are less significant.

(6.) While the realised spread estimators have been criticised (see Wang et al. 1994, fn. 6), we replicated the Wang et al. (1994) estimator for the Sydney Futures Exchange. Our findings were identical to Wang et al. (1994) suggesting that the behaviour of the markets are similar, and the differences in findings are driven by the actual metrics.

(7.) See Ma, Peterson and Sears (1992) and Wang et al. (1994).

(8.) The SPI futures contract is based on the All Ordinaries Index, which is constructed from approximately 300 stocks traded on the Australian Stock Exchange (ASX). See Aitken and Frino (1996) for a more complete description of equities trading on the ASX.

(9.) Stocks on the ASX do not open simultaneously. Rather, they are grouped according to their alphabetical ranking, and each group is opened randomly at different times between 9.59 a.m. and 10.09 a.m.

(10.) Lunchtime trading was introduced on 9 May 1994 for the SPI contract, and removed on 1 October 1994.

(Date of receipt of final typescript: August 1998 Accepted by Tom Smith, Area Editor.)

References

Aitken, M., Brown, P. & Walter, T. 1994, Intraday patterns in returns, volume, volatility and trading frequency on SEATS, Working paper, SIRCA, Sydney.

Aitken, M.J. & Frino, A. 1996, `The determinants of market bid ask spreads on the Australian Stock Exchange: Cross-sectional analysis', Accounting and Finance, vol. 36, pp. 51-64.

Bagehot, W. 1971, `The only game in town', Financial Analysts Journal, vol. 27, 12-22.

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, vol. 16, pp. 451-89.

Chan, K. 1992, `A further analysis of the lead-lag relationship between the cash market and stock index futures market', Review of Financial Studies, vol. 5, pp. 123-52.

Chan, K.C., Christie, W.G. & Schultz, P.H. 1995, `Market structure and the intraday pattern of bid ask spreads for NASDAQ securities', Journal of Business, vol. 68, pp. 35-60.

Chan, K.C., Chung, Y.P. & Johnson, H. 1995, `The intraday behaviour of bid-ask spreads for NYSE stocks and CBOE options', Journal of Financial and Quantitative Analysis, vol. 30, pp. 329-46.

Copeland, T.E. & Galai, D. 1983, `Information effects of the bid-ask spread', Journal of Finance, vol. 38, pp. 1457-69.

Ederington, L. & Lee, J. 1993, `How markets process information: New releases and volatility', Journal of Finance, vol. 48, pp. 1161-91.

Fleming, J., Ostdiek, B. & Whaley, R.E. 1996, `Trading costs and the relative rates of price discovery in stock, futures, and option markets', Journal of Futures Markets, vol. 16, pp. 353-87.

Foster, F. & Viswanathan, S. 1993, `Variations in trading volume, return volatility and trading costs: Evidence on recent price formation models', Journal Finance, vol. 48, pp. 187-211.

Foster, F. & Viswanathan, S. 1994, `Strategic trading with asymmetrically informed traders and long-lived information', Journal of Financial and Quantitative Analysis, vol. 29, pp. 499-518.

Hansen, L. 1982, `Large sample properties of generalised method of moment estimator', Econometrica, vol. 50, pp. 1029-54.

Ho, T. & Macris, R.G. 1985, `Dealer market structure and performance', ch. 3, in Market Making and the Changing Structure of the Securities Industry, eds Y. Amihud, T.S.Y. Ho & R.A. Schwartz, Lexington Books, New York.

Ho, T. & Stoll, H.R. 1980, `On dealer markets under competition', Journal of Finance, vol. 35, pp. 259-67.

Ho, T. & Stoll, H.R. 1983, `The dynamics of dealer markets under competition', Journal of Finance, vol. 38, pp. 1053-74.

Jennings, R. 1994, `Intraday changes in target firms' share price and bid ask quotes around takeover announcements', Review of Financial Research, vol. 17, pp. 255-70.

Ma, C.K., Peterson, R.L. & Sears, R.S. 1992, `Trading noise, adverse selection and the intraday bid ask spreads in futures markets', Journal of Futures Markets, vol. 12, pp. 519-38.

Madhavan, A. 1992, `Trading mechanisms in securities markets', Journal of Finance, vol. 47, pp. 607-41.

Manaster, S. & Mann, S.C. 1996, `Life in the pits: Competitive market making and inventory control', Review of Financial Studies, vol. 9, pp. 953-75.

McInish, T.H. & Wood, R.A. 1992, `An analysis of intraday patterns in bid/ask spreads for NYSE stocks', Journal of Finance, vol. 47, pp. 753-64.

Newey, W. & West, K. 1994, `Automatic lag selection in covariance matrix estimation', Review of Economic Studies, vol. 61, pp. 631-53.

Smith, T. & Whaley, R.E. 1994, `Estimating the effective bid/ask spread from time and sales data', Journal of Futures Markets, vol. 14, pp. 437-55.

Stoll, H.R. 1978, `The supply of dealer services in securities markets', Journal of Finance, vol. 33, pp. 1133-51.

Venkatesh, P. 1992, `Empirical evidence on the impact of the bid ask spread on the characteristics of CRSP daily returns', Journal of Financial Research, vol. 15, pp. 113-25.

Wang, G.H.K., Michalski, R.J., Jordon, J.V. & Moriarty, E.J. 1994, `An intraday analysis of bid ask spreads and price volatility in the S&P 500 index futures market', The Journal of Futures Markets, vol. 14, pp. 837-59.

Appendix

Institutional Setting

Trading on the Sydney Futures Exchange (SFE) is conducted on a trading floor through continuous open outcry, and is analogous to trading on the Chicago Mercantile Exchange (CME), which has been analysed in previous research.(7) There are two main types of traders; Floor Members and Local Members. A floor member has full access to the trading floor and is permitted to trade as either a principal (on their own account) or an agent (for clients) in accordance with the Business Rules of the SFE. Locals typically trade on their own account. However, they are permitted to execute orders for other brokers on a `give-up' basis. In doing give-up business, locals temporarily suspend their principal trading activity (e.g. for one day) and trade on behalf of other brokers for a fee.

Each pit is occupied by a `Price Reporter', who is an exchange official. Price reporters use microphones to communicate with Price Report System Operators located on a catwalk above the trading floor. Price reporters communicate trade information to the Operator, including trade prices and volumes, as well as bid and ask quotes. The Operators then enter the trade information onto computer terminals, which automatically assign a time stamp to each trade. This data collection system is referred to as the Price Reporting System, and the data is sold on-line in real-time to quote vendors. The format of the data sold to quote vendors conforms to Inter-Exchange Technical Committee (ITC) Standards, and is frequently referred to as ITC data. ITC data is comparable to US data, more commonly referred to as `time and sales' data. However, time and sales data reports the prices of price changing trades only. Hence, the SFE's ITC data provides a unique opportunity to directly measure quoted bid-ask spreads.

Floor trading on the SFE commences at 8.30 a.m. and ends at 4.30 p.m. for 3-year bond futures, 10-year bond futures and 90-day bank accepted bill futures. Trading of the Share Price Index (SPI) futures commences at 9.50 a.m. and ends at 4.10 p.m.(8) In contrast, the stocks from which the index is constructed are traded on the ASX from 10.00 a.m. to 4 00 p.m.(9) Unlike overseas exchanges and the Australian Stock Exchange, trading on the SFE breaks for lunch between 12.30 p.m. and 2.00 p.m.(10)

Alex Frino and Matthew Duffy, Department of Finance, The University of Sydney, Sydney NSW 2006.

Max Stevenson, School of Finance and Economics, UTS, Broadway NSW 2007.

This research is funded by an ARC Collaborative Grant (No. C59700105) involving the Sydney Futures Exchange. The authors thank Nic McGilvray for programming support, Mike Aitken and Tom McInish for helpful comments and participants at a workshop held by the Research and Development Division of the Sydney Futures Exchange.3
COPYRIGHT 1998 Australian Graduate School Of Management
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Duffy, Matthew
Publication:Australian Journal of Management
Date:Dec 1, 1998
Words:7559
Previous Article:The Role of R&D Capitalisations in Firm Valuation and Performance Measurement.
Next Article:The Pricing of Low Exercise Price Options.
Topics:


Related Articles
The pricing of Australian index futures contracts with taxes and transaction costs.
Wall Street turns its back on Vencor, Sun. NewCare, and ALC.
A fraction at a time.
Recovering from 'the slings and arrows of outrageous fortune'.
NEXA TECHNOLOGIES AND TSX DATALINK FORM PARTNERSHIP.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters