# An empirical analysis of the impact of stock index futures trading on securities dealers' inventory risk in the NASDAQ market.

Abstract

Stock index futures contracts (SIFCs) were developed in part to allow equity investors to conveniently hedge portfolio risks. Therefore, we may expect to observe smaller bid/ask spreads among NASDAQ equities following the introduction of trading in SIFCs. The potential transactions cost reduction results from the enhanced ability of dealers to hedge their inventory risk. Accordingly, we analyze the daily returns and bid/ask spreads of all CRSP-listed NASDAQ stocks for a 1 0-year period surrounding the introduction of three SIFCs in order to determine whether the introduction of these contracts has affected transactions costs in the NASDAQ equity market. Our results indicate that bid/ask spreads narrowed significantly following the introduction of stock index futures trading. Moreover, the behavior of the spreads leads us to the conclusion that risk reduction through cost-effective hedging via stock index futures trading may have been the cause. [C] 2002 Published by Elsevier Science Inc.

JEL classification: G13; G14

Keywords: Stock index; Futures; NASDAQ; Systematic risk; Bid/ask spread

1. Introduction

Stock index futures contracts (SIFCs) are popular among investors because they offer a low-cost, convenient way to hedge portfolio risk (Weiner, 1981). Financial economists have demonstrated that maximum risk reduction may be obtained by employing a hedge ratio equal to a portfolio's beta (see, for example, Ederington, 1979; Figlewski & Kon, 1982). Empirical evidence suggests that hedging with SIFC may significantly enhance an investor's reward-to-risk ratio (Figlewski, 1985; Junkus & Lee, 1985; Nordhauser, 1984). While previous research has dealt with the benefits of SIFC trading to investors, the effect of SIEC trading on dealer costs has not been adequately addressed.

Bid/ask spreads observed in the NASDAQ market reflect the dealer's inventory holding costs, order processing costs, and the cost of trading against informed traders. SIFC trading may affect any of these components by affecting market liquidity and price volatility. The reader is directed to Coughenour and Shastri (1999) for a review of the literature on the determinants of dealer bid/ask spreads. We investigate whether the advent of SIFC trading allowed dealers to more efficiently manage the price risk associated with their trading inventories by analyzing the behavior of bid/ask spreads on NASDAQ stocks for a 10-year period surrounding the introduction of SIFC.

Our analysis reveals three findings. First, we find that a stock's bid/ask spread is positively related to its systematic risk. Moreover, the sensitivity of the spread to changes in portfolio risk is reduced in the post-SIFC trading period. We interpret this result as evidence that the introduction of SIFC trading has allowed security dealers to protect themselves against adverse market movements. Second, we observe that bid/ask spreads are lower in the post-SIFC period, even after controlling for possible structural changes caused by SIFC trading. We interpret this result as evidence of the enhanced availability of cost-effective hedging opportunities. Finally, we find that the reduced sensitivity of spreads to changes in portfolio risk observed in the post-SIFC trading period is greater in magnitude and statistical significance during periods of high market volatility and for securities with high systematic risk. We interpret this result as evidence that rational utility-maximizing market makers concentrat e their hedging efforts on higher risk stocks and during periods of greater market volatility.

The plan of the paper is as follows. In Section 2, we summarize the relevant extant literature relating to equity market microstructure, observed bid/ask spreads, and stock index futures trading. Section 3 describes briefly some important aspects of hedging with stock index futures, along with a brief general description of our empirical tests. Section 4 presents descriptions of our sample, methodology, and results. Our results are divided into three parts: (1) comparisons of bid/ask spreads and various risk measures between the pre- and post-SIFC periods, (2) regression analyses of bid/ask spreads against measures of portfolio risk and dummy variables relating to SIFC introduction, and (3) regression analyses of various subsamples designed to determine whether observed structural changes in the post-SIFC period are related to market volatility and portfolio risk. We present our conclusions in Section 5.

2. Relevant research

Most of the extant research on the effect of derivatives trading on the microstructure of the underlying security market examines directly the relationship between the derivative and the underlying market. Black (1975) and Cox and Rubinstein (1985) posit that informed traders prefer to trade in the options markets for three reasons. First, trading options effectively allows the investor to lever his/her position well in excess of the allowable margin rate prevailing in the equity market. Second, more favorable borrowing and lending rates are imbedded in options prices. (2) Third, short-sale restrictions can be circumvented in the options market. (3) As a result of these considerations, we believe that the introduction of options trading may cause bid/ask spreads in the underlying market to narrow, as informed traders migrate to the options markets. This migration causes a reduction in the adverse selection component of the observed bid/ask spread.

An organized options market increases the returns from investments in information production. This provides greater incentive to produce more information for firms whose stocks are optioned. For example, Skinner (1990) observes a significant increase in the number of analysts following a given stock following the initiation of options trading in that stock. Enhanced information processing efficiency in the options market may translate into enhanced informational efficiency in the market for the underlying stock, given the short life of arbitrage opportunities between the two markets. (4) This argument motivates the hypothesis that the markets for optioned stocks may be characterized by increased liquidity, lower bid/ask spreads, and rapid adjustment to new information. Not surprisingly, financial economists have observed these developments for optioned stocks (see Conrad, 1989; Kumar, Sarin, & Shastri, 1998; Skinner, 1990).

Some analysts disagree with the notion that SIFCs enhance market efficiency. This contingent believes that index arbitrage programs may actually transmit volatility from the index futures market to the underlying equities market. (5) They cite the stock market crash of October 1987 as an example in which price pressure originating in the index futures market resulted in a collapse in stock prices. In response to these allegations, Miller (1991) argues that SIFC trading only appeared to lead the market because prices of many stocks in the index were not updated from the previous day's close. To date, the impact of SIFC trading on the underlying stock market is an unresolved empirical question.

Although options on the underlying stocks may be used in order to hedge a portfolio, SIFC may be the preferred vehicle for reasons of convenience, low cost, and high liquidity. These characteristics may be especially important for security dealers because portfolio risk is hypothesized to be a major determinant of the dealer's bid/ask spread on a given stock (see Demsetz, 1968; Stoll, 1978a; Tinic, 1972). For example, Stoll (1978a) hypothesizes that both systematic and unsystematic risk should matter to the securities dealer because the marginal position of a dealer at any point in time is undiversified. His empirical findings (Stoll, 1978b) support this hypothesis. In contrast to Stoll's research, Benston and Hagerman (1974) argue that systematic risk should not affect the spread because the dealer's expected return from holding the security compensates him/her for all relevant risks. Their empirical work supports their position. More recently, Tripathy and Peterson (1991) extend empirical testing of the re lationship between risk and the bid/ask spread to account for possible intertemporal changes in systematic risk, and they conclude that systematic risk has no effect on the bid/ask spreads of OTC stocks. We speculate that differences in systematic risk impacts between Stoll (1978b) and Tripathy and Peterson may be due to the fact that Stoll's work predates the advent of SIFC, while Tripathy and Peterson's data follows introduction of SIFC trading.

3. Hedging dealer inventory with stock index futures

The simplest hedging strategy using SIFCs involves purchasing the number of contracts given by:

[Portfolio Value x Portfolio Beta]/[Current Stock Index Value x 500].

Assuming that dealers are more inclined to carry inventory that is easily hedged, we hypothesize that new hedging opportunities brought about by the advent of SIFCs will reduce dealer inventory holding costs. Competition among dealers will induce them to pass at least part of the cost savings onto investors in the form of narrowed bid/ask spreads. Moreover, we believe that even a partial hedge may significantly lower the dealer's inventory holding costs. (6)

As a result of the above considerations, we test empirically the hypothesis that SIFC trading is associated with reduced bid/ask spreads and with decreased sensitivity of bid/ask spreads to changes in portfolio risk. In addition, we explore the possibility that factors may exist that effectively constrain the market maker's use of SIFC. An example of such a factor is the round-trip cost of transacting. Walter and Welles (1988) estimate typical round-trip transactions costs, including internal order clearing, brokerage, and market spread costs, to be approximately US$55. Although this may represent a transactions cost savings relative to alternative hedging techniques, it is still probably unrealistic to expect market makers to hedge indiscriminately across all stocks.

Inspired by Walter and Welles (1988), we examine whether security dealers exercise timing and selectivity in their hedging activities. The terms timing and selectively are traditionally used to describe investment performance criteria (Admati, Bhattacharya, Pfleiderer, & Ross, 1986). In the context of this research, we refer to "timing" as the ability of dealers to hedge inventory risk due to market swings. Walter and Welles (1988) report that specialist firms produce better results when they hedge against fundamental shifts in the market return distribution rather than against day-to-day fluctuations. We refer to "selectivity" as the ability of dealers to pick a subset of stocks to hedge so as to maximize the benefits of hedging. In light of these considerations, we hypothesize that bid/ask spread reductions due to the effect of hedging SIFCs on dealer inventory risk will be most pronounced during periods of high market volatility and for high-beta stocks.

4. Data, methodology, and results

4.1. Sample description

Table 1 lists the dates and exchanges upon which organized trading in four major SIFCs was introduced. We define the SIFC trading introduction period as the 51 trading days between February 24, 1982 and May 6, 1982. This time interval spans the introduction of the first three SIFCs. (7) We further define pre- and post-SWC introduction periods as the 5-year period preceding February 24, 1982, and the 5-year period following May 6, 1982, respectively. We analyze a sample consisting of all 2495 stocks trading on the NASDAQ system for which daily returns and bid/ask spreads are available from the Center for Research in Securities Prices NASDAQ database for a 10-year period surrounding the SIFC introduction period described above.

4.2. Pre- and the post-SIFC introduction period comparison

Table 2 presents descriptive statistics and mean difference t tests for percentage bid/ask spreads, betas, systematic risk measures, and total risk measures for the pre- and post-SIFC introduction periods. Statistical tests reveal that the mean percentage bid/ask spread is smaller following SIFC introduction at conventional significance levels. Moreover, the means of the three risk measures are significantly higher following SIFC introduction. These findings are especially interesting in light of the fact that bid/ask spreads are thought to be positively related to risk (see Stoll, 1978b). We conclude from these findings that observed bid/ask spread reductions following SIFC trading are likely not due to lower intrinsic risk in stock prices.

Next, we partition our sample into quintiles along three dimensions, namely, market value, return volatility, and beta. We then perform a series of mean difference t tests. Table 3, Panel A shows quintile means before and after SIFC introduction, along with the results of t tests by market value quintile. Table 3, Panel B reports analogous statistics by return volatility quintile. Table 3, Panel C reports analogous statistics by beta quintile. We find that mean bid/ask spreads increase (decrease) monotonically with market value (return volatility). Moreover, the collective evidence from Table 3 suggests that (excluding the lowest quintiles for return volatility and beta) the bid/ask spread reduction observed in the post-SIFC introduction period is significant at the 10% level or better in all cases, regardless of how the sample is partitioned. We interpret these findings as evidence that the potential benefits to SIFC hedging span equities of all sizes and risks.

4.3. Bid/ask spreads and portfolio risk

We test whether there exists a positive relationship between observed bid/ask spreads and portfolio risk, and further, whether this relationship is less elastic in the period following SIFC introduction. These tests are based upon the following regression model:

SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i] RIS[K.sub.i] + [[epsilon].sub.i] (1)

where subscript i = index value representing stock I, SPREA[D.sub.i] = [As[k.sub.i] - Bi[d.sub.i]]/[(As[k.sub.i] + Bi[d.sub.i])/2], RIS[K.sub.i] = portfolio risk of holding stock i, SIF[C.sub.i] = 1 if the observation is drawn from the post-SIFC introduction period, and zero otherwise, SIF[C.sub.i].RIS[K.sub.i] = dummy variable for the regression slope associated with RIS[K.sub.i], and [[epsilon].sub.i] = error term, assumed to meet OLS specifications.

We estimate Eq. (1) using a pooled sample drawn from the pre- and post-SIFC trading periods. (8) SPREA[D.sub.i] is a 5-year average value, with the 5-year average computed from data prior to (after) SIFC introduction if the observation is drawn from the pre- (post-) SIFC introduction period. Two alternative measures are computed to represent RIS[K.sub.i]. First, we define RIS[K.sub.i] as [[beta].sub.i.sup.2] x [[sigma].sub.m.sup.2], where [[beta].sub.i] is the stock's OLS beta estimated from a market model regression, and [[sigma].sub.m.sup.2] is the variance of NASDAQ market returns. We define RIS[K.sub.i] alternatively as the stock return variance, [[sigma].sub.i.sup.2]. We estimate regressions using the second RIS[K.sub.i] definition in light of well-known problems in estimating beta (i.e., infrequent trading, nonsynchronous stock and index returns, and market model nonstationarities). Moreover, Stoll (1978b) posits that both systematic risk and total risk affect the bid/ask spread.

SIF[C.sub.i] and SIF[C.sub.i]*RIS[K.sub.i] are intercept and slope dummy variables designed to test for structural changes in market microstructure associated with the introduction of SIFCs. We predict that the coefficient associated with SIF[C.sub.i] will be significantly negative because we believe that the advent of SIFCs should be associated with smaller bid/ask spreads for reasons outlined earlier. Furthermore, we predict that the slope dummy variable SIF[C.sub.i]*RIS[K.sub.i] will be significantly negative because we believe that SIFC trading to some degree immunizes dealers from the effects of portfolio risk, thereby reducing the elasticity of the spread with respect to risk.

We report full-sample results for estimation of Eq. (1) in Table 4. We report quintile results for regressions estimated from subsamples disaggregated by market value, return volatility, and beta in Table 5, Panels A through C, respectively. For the sake of brevity, we report regression results only for RIS[K.sub.i] proxied by total risk in Table 5. Based upon our interpretation of the coefficients and significance tests reported in Tables 4 and 5, we present three salient findings.

First, we observe from Tables 4 and 5 that, in general, portfolio risk is positively related to the bid/ask spread. This implies that security dealers may be affected by portfolio risk, and thus there exists the potential for dealers to benefit from hedging opportunities. Second, based upon our interpretation of the significant negative coefficients associated with the slope dummy variable SIFC.RISK reported in Tables 4 and 5, Panel C, the sensitivity of bid/ask spreads with respect to portfolio risk is reduced in the post-SIFC trading period. We interpret this finding as evidence that security dealers reduced their required compensation for market-related risk following the introduction of SIFCs. Table 5, Panels A and B report mixed results. Finally, the significantly negative coefficients for variable SIFC reported in Tables 4 and 5, Panel C suggest that there is a secular decline in the percentage bid/ask spread in the post-SIFC period. We speculate that this decline may have resulted from increased tradin g volume (and liquidity), and/or from reduced adverse selection costs in the NASDAQ market following SIFC introduction. Again, we observed mixed results for variable SIFC in Table 5, Panels A and B. (9)

We would like to include market microstructure variables, such as trading volume and stock price, in our estimated regressions. Unfortunately, trading volume data are not available for stocks in the pre-SIFC trading period. Nevertheless, we attempt to incorporate volume effects in our models by dividing the sample into volume quintiles using the available post-SIFC volume data. We then estimate regression models by quintile. We anticipate that volume (by quintile) will interact with the relationship between the bid/ask spread and the explanatory variables. We also include the stock price, with the expectation that it is negatively correlated with the percentage bid/ask spread. Finally, as an alternative to the stock price, we include market capitalization in the regression specification. We present the results from these attempts to incorporate general microstructure variables in the regressions in Table 6.

Results from the experiments with market microstructure variables yield three observations. First, as anticipated, stock prices and, alternatively, market capitalization values are significantly negatively related to the percentage bid/ask spread, with the exception of market capitalization in the case of the highest volume quintile. Second, the inclusion of the microstructure variables improves the overall significance and explanatory power of the estimated regressions. Finally, the slope and intercept dummy coefficients are uniformly significantly negative through the middle three quartiles in Table 6, Panels A and B.

Overall, our regression models reveal three findings. First, percentage bid/ask spreads are lower in the post-SIFC period. Second, while the percentage bid/ask spread is negatively related to portfolio risk generally, the percentage spread is less sensitive to changes in portfolio risk in the post-SIFC period. Finally, the observed relationships are robust with respect to controls for market capitalization, return volatility, beta, stock price, and trading volume.

4.4. Timing and selectivity effects

Our final set of tests is designed to determine whether the magnitude of bid/ask spread changes between pre- and post-SIFC trading periods is greater during periods of high market volatility and for high beta stocks, as discussed in Section 3. For the analysis with respect to timing, we select four quarters from our pre- and post-SIFC intervals representing four regimes: (1) high volatility, pre-SIFC, (2) low volatility, pre-SIFC, (3) high volatility, post-SIFC, and (4) low volatility, post-SIFC. We then estimate two regression models, one corresponding to low volatility markets and the other corresponding to high volatility markets. For the analysis with respect to selectivity, we estimate two regression models, one for below median beta stocks and the other for above median beta stocks.

In order to implement the timing tests, we use the CRSP NASDAQ database to construct a series of rolling 30-day return variances for returns on the NASDAQ index. The quarterly averages from this rolling series are illustrated in Fig. 1 for the pre-SIFC period and in Fig. 2 for the post-SIFC period. For the pre-SIFC period, 1980 II and 1979 III are the quarters associated with the highest and lowest index return variances, respectively. For the post-SIFC period, the corresponding maximum and minimum average index variance quarters are 1982 IV and 1985 II, respectively. (10)

Table 7, Panel A presents regression estimates for Eq. (1) for data from pre- and post-SIFC high variance quarters, and also for data from pre- and post-SIFC low variance quarters. All regression coefficients reported in Table 7, Panel A are significant at the 10% level or better, except for the slope dummy variable in the low variance model. All other estimated coefficients are similar to our original results. We interpret these results as evidence that hedging opportunities afforded by SIFC trading may be more valuable to dealers and investors during periods of high market volatility than is the case during periods of low market volatility. (11)

Table 7, Panel B presents regression estimates for Eq. (1) for data from pre- and post-SIFC high beta stocks, and also for data from pre- and post-SIFC low beta stocks. As was the case in Table 7, Panel A, all regression coefficients reported in Table 7, Panel B are significant at the 10% level or better, except for the slope dummy variable in the low beta model. All other estimated coefficients are similar to our original results. We interpret these results as evidence that the hedging opportunities afforded by SIFC trading may be most valuable for hedging higher beta stocks.

Before concluding the current set of tests, we conduct an alternative test of the selectivity hypothesis. We estimate the following regression (Eq. (2)) in order to determine whether high beta stocks experienced a larger decline in bid/ask spreads than low beta stocks:

CHSPREAD = -0.02[6.sub.(-7.5)] - 0.008[2.sub.(-2.6)] BETA (2)

where CHSPREAD = the difference between 5-year average bid/ask spreads for each stock estimated over the pre- and post-SIFC trading period, BETA = the OLS estimate of beta for the same stock estimated using 5 years of monthly returns over the pre-SIFC trading period. We interpret the statistically significant negative slope coefficient as evidence to support the hypothesis that higher beta stocks exhibit a larger decline in bid/ask spreads in the post-SIFC trading period. Overall, the results from our timing and selectivity tests are consistent with the notion that dealers and market makers either choose, or are able, to more effectively hedge inventory risk with SIFCs during periods of high market volatility and for high beta stocks.

5. Conclusion

We analyze the behavior of bid/ask spreads on NASDAQ stocks before and after the introduction of stock index futures trading. Our analysis reveals the following findings. First, we find that a stock's bid/ask spread is positively related to its systematic risk. We interpret this finding as evidence that dealers "price" portfolio risk. Dealers may therefore benefit from hedging this risk by trading SIFCs. Moreover, we find that the sensitivity of spreads to changes in portfolio risk is reduced in the post-SIFC trading period. We interpret this finding as evidence that the introduction of SIFC trading has allowed security dealers to reduce their portfolio risk.

As a result, dealers are able to reduce bid/ask spreads in the market. Second, we document an overall decline in percentage bid/ask spreads in the post-SIFC trading period, even after controlling for possible structural changes caused by SIFC trading. We attribute this reduction in spreads to the increased availability of cost-effective hedging opportunities provided by SIFCs. (12)

Since hedging costs represent a tangible expense that must be borne by market participants desiring to hedge their risks, we also examine whether dealers may benefit from timing their hedging activities, and/or selectively employing hedging strategies primarily for high systematic risk stocks. While we find that the positive relationship between bid/ask spreads and portfolio risk is lessened following SIFC introduction, we find that this phenomenon is more prominent during periods of high market volatility. We also find that higher beta stocks exhibit a larger reduction in bid/ask spreads in the post-SIFC trading period than lower beta stocks. We interpret these findings as evidence consistent with the notion that rational utility-maximizing traders concentrate their hedging efforts and resources on higher risk stocks and during periods of greater market volatility.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

Acknowledgments

We would like to thank Peggy Swanson, J. David Diltz, two anonymous reviewers, and the editor of this journal for their constructive suggestions. Any remaining errors are our own.

(1.) Tel.: +1-940-565-3045.

(2.) To the extent that the borrowing and lending rates are applicable to large market participants who set option prices on the margin, these rates can be favorable to the typical options trader.

(3.) Under the current rules, a short sale can be executed only when the previous nonzero price change is positive. Also, use of options market avoids loss of frill proceed of short-sale transaction in the stock market.

(4.) Stephan and Whaley (1990) report that price adjustments across the stock and options markets are accomplished in 15 minutes. In the case of block trades, Kumar, Sarin, and Shastri (1992) observe that significant lead--lag relationship of returns between the two occur within only a 30-minute interval.

(5.) See Miller (1991) for a detailed discussion of this issue.

(6.) As Figlewski and Kon (1982) observe, only a stock portfolio whose composition is identical to the index can be perfectly hedged. Stock index futures, which are essentially cross-hedges, will therefore provide only incomplete hedges against systematic risk in the marker's portfolio.

(7.) We do not consider the incremental effect of the fourth SIFC that was introduced July 23, 1984. We exclude the fourth SIFC from our consideration because we believe that our current 4-month introduction period would introduce smaller error in our statistical estimates than 30-month window that would be required if we were to include the fourth SIFC in our testing. Moreover, we trade the other three contracts prior to July 23, 1984.

(8.) We recognize that this regression specification is incomplete in the sense that it does not account for other bid/ask spread determinants, such as volume, number of dealers, and stock price. We incorporate some of these variables in subsequent testing.

(9.) In order to determine the stability of cross-sectional regression coefficients, we repeated our regression estimation using two and four intervals instead of the year interval reported in the body of the paper. We did not find any qualitative difference in the sign, magnitude, or regression coefficients estimated from these alternate windows.

(10.) We observe that the market variance levels in the selected high and low quarters from the pre-index periods are noticeably higher than the corresponding quarters in the post-index period. To preserve our results from contamination by this bias, we select quarters with approximately equal high and low levels of market variance from both periods:

(11.) We must note, however, that if market volatility interacts systematically with the spread determinants in our sample, then the results presented in Table 7, Panel A must be regarded as inconclusive.

(12.) We acknowledge, however, that observed changes in bid/ask spreads are also consistent wit

References

Admati, A. R., Bhattacharya, S., Pfleiderer, P., & Ross, S. A. (1986). On timing and selectivity. Journal of Finance, 41, 715-729.

Benston, G. G., & Hagerman, R. L. (1974). Determinants of bid--ask spreads in the over-the-counter market. Journal of Financial Economics, 2, 353-364.

Black, F. (1975). Facts and fantasy in the use of options. Financial Analysts Journal, 31, 61-72.

Conrad, J. (1989). The price effect of options introduction. Journal of Finance, 44, 487-498.

Coughenour, J., & Shastri, K. (1999). Symposium on market microstructure: a review of empirical literature. Financial Review, 34, 1-28.

Cox, J. C., & Rubinstein, M. (1985). Options markets. Englewood Cliffs, NJ: Prentice Hall.

Demsetz, H. (1968). The cost of transacting. Quarterly Journal of Economics, 32, 33-53.

Ederington, L. (1979). The hedging performance of the new futures market. Journal of Finance, 34, 157-170.

Figlewski, S. (1985). Hedging with stock index futures: theory and application in a new market. Journal of Futures Market, 5, 183-199.

Figlewski, S., & Kon, S. J. (1982). Portfolio management with stock index futures. Financial Analysts Journal, 38, 52-60.

Junkus, J. C., & Lee, C. F. (1985). Use of three stock index futures in hedging decisions. Journal of Futures Market, 5, 201-222.

Kumar, R., Sarin, A., & Shastri, K. (1992). The behaviour of option price around large block transactions in the underlying security. Journal of Finance, 47, 879-890.

Kumar, R., Sarin, A., & Shastri, K. (1998). The impact of options trading in the market quality of the underlying security: an empirical analysis. Journal of Finance, 53, 717-732.

Miller, M. H. (1991). Financial innovations and market volatility. Cambridge, MA: Blackwell Publishers.

Nordhauser, F. (1984). Using stock index futures to reduce market risk. Journal of Portfolio Management, 10, 56-62.

Skinner, D. J. (1990). Options markets and the information content of accounting earnings releases. Journal of Accounting and Economics, 13, 191-211.

Stephan, J., & Whaley, R. (1990). Intraday price change and trading volume relations in the stock and stock options markets. Journal of Finance, 55, 191-220.

Stoll, H. (1978a). The supply of dealer service to the securities markets. Journal of Finance, 33, 1133-1151.

Stoll, H. (1978b). The pricing of security dealer's service: an empirical study of NASDAQ stocks. Journal of Finance, 33, 1153-1172.

Tinic, S. M. (1972). The economics of liquidity services. Quarterly Journal of Economics, 36, 79-93.

Tripathy, N., & Peterson, R. L. (1991). Risk adjustment costs and securities dealer's bid-ask spreads. Journal of Financial Service Research, 5, 131-142.

Walter, R. C., & Welles, E. C. (1988). Hedging the market risk in the specialist portfolio. Presented at the Financial Management Association Annual Meeting, New Orleans.

Weiner, N. S. (1981). The hedging rationale for a stock index futures contract. Journal of Futures Market, 1,59-76.

Salil K. Sarkar (a) *, Niranjan Tripathy (b,1)

(a) Finance and Real Estate Departmnet, University of Texas at Arlington, Box 19499, Arlington, TX 76019, USA

(b) Departmnet of Finanace, Insurance, Real Estate and Law, University of North Texas, Denton, TX 76203, USA

* Corresponding author. Tel.: +1-817-272-3836; fax: +1-817-272-2252.

E-mail addresses: sarkar@uta.edu (S.K. Sarkar), tripathy@unt.edu (N. Tripathy).

Stock index futures contracts (SIFCs) were developed in part to allow equity investors to conveniently hedge portfolio risks. Therefore, we may expect to observe smaller bid/ask spreads among NASDAQ equities following the introduction of trading in SIFCs. The potential transactions cost reduction results from the enhanced ability of dealers to hedge their inventory risk. Accordingly, we analyze the daily returns and bid/ask spreads of all CRSP-listed NASDAQ stocks for a 1 0-year period surrounding the introduction of three SIFCs in order to determine whether the introduction of these contracts has affected transactions costs in the NASDAQ equity market. Our results indicate that bid/ask spreads narrowed significantly following the introduction of stock index futures trading. Moreover, the behavior of the spreads leads us to the conclusion that risk reduction through cost-effective hedging via stock index futures trading may have been the cause. [C] 2002 Published by Elsevier Science Inc.

JEL classification: G13; G14

Keywords: Stock index; Futures; NASDAQ; Systematic risk; Bid/ask spread

1. Introduction

Stock index futures contracts (SIFCs) are popular among investors because they offer a low-cost, convenient way to hedge portfolio risk (Weiner, 1981). Financial economists have demonstrated that maximum risk reduction may be obtained by employing a hedge ratio equal to a portfolio's beta (see, for example, Ederington, 1979; Figlewski & Kon, 1982). Empirical evidence suggests that hedging with SIFC may significantly enhance an investor's reward-to-risk ratio (Figlewski, 1985; Junkus & Lee, 1985; Nordhauser, 1984). While previous research has dealt with the benefits of SIFC trading to investors, the effect of SIEC trading on dealer costs has not been adequately addressed.

Bid/ask spreads observed in the NASDAQ market reflect the dealer's inventory holding costs, order processing costs, and the cost of trading against informed traders. SIFC trading may affect any of these components by affecting market liquidity and price volatility. The reader is directed to Coughenour and Shastri (1999) for a review of the literature on the determinants of dealer bid/ask spreads. We investigate whether the advent of SIFC trading allowed dealers to more efficiently manage the price risk associated with their trading inventories by analyzing the behavior of bid/ask spreads on NASDAQ stocks for a 10-year period surrounding the introduction of SIFC.

Our analysis reveals three findings. First, we find that a stock's bid/ask spread is positively related to its systematic risk. Moreover, the sensitivity of the spread to changes in portfolio risk is reduced in the post-SIFC trading period. We interpret this result as evidence that the introduction of SIFC trading has allowed security dealers to protect themselves against adverse market movements. Second, we observe that bid/ask spreads are lower in the post-SIFC period, even after controlling for possible structural changes caused by SIFC trading. We interpret this result as evidence of the enhanced availability of cost-effective hedging opportunities. Finally, we find that the reduced sensitivity of spreads to changes in portfolio risk observed in the post-SIFC trading period is greater in magnitude and statistical significance during periods of high market volatility and for securities with high systematic risk. We interpret this result as evidence that rational utility-maximizing market makers concentrat e their hedging efforts on higher risk stocks and during periods of greater market volatility.

The plan of the paper is as follows. In Section 2, we summarize the relevant extant literature relating to equity market microstructure, observed bid/ask spreads, and stock index futures trading. Section 3 describes briefly some important aspects of hedging with stock index futures, along with a brief general description of our empirical tests. Section 4 presents descriptions of our sample, methodology, and results. Our results are divided into three parts: (1) comparisons of bid/ask spreads and various risk measures between the pre- and post-SIFC periods, (2) regression analyses of bid/ask spreads against measures of portfolio risk and dummy variables relating to SIFC introduction, and (3) regression analyses of various subsamples designed to determine whether observed structural changes in the post-SIFC period are related to market volatility and portfolio risk. We present our conclusions in Section 5.

2. Relevant research

Most of the extant research on the effect of derivatives trading on the microstructure of the underlying security market examines directly the relationship between the derivative and the underlying market. Black (1975) and Cox and Rubinstein (1985) posit that informed traders prefer to trade in the options markets for three reasons. First, trading options effectively allows the investor to lever his/her position well in excess of the allowable margin rate prevailing in the equity market. Second, more favorable borrowing and lending rates are imbedded in options prices. (2) Third, short-sale restrictions can be circumvented in the options market. (3) As a result of these considerations, we believe that the introduction of options trading may cause bid/ask spreads in the underlying market to narrow, as informed traders migrate to the options markets. This migration causes a reduction in the adverse selection component of the observed bid/ask spread.

An organized options market increases the returns from investments in information production. This provides greater incentive to produce more information for firms whose stocks are optioned. For example, Skinner (1990) observes a significant increase in the number of analysts following a given stock following the initiation of options trading in that stock. Enhanced information processing efficiency in the options market may translate into enhanced informational efficiency in the market for the underlying stock, given the short life of arbitrage opportunities between the two markets. (4) This argument motivates the hypothesis that the markets for optioned stocks may be characterized by increased liquidity, lower bid/ask spreads, and rapid adjustment to new information. Not surprisingly, financial economists have observed these developments for optioned stocks (see Conrad, 1989; Kumar, Sarin, & Shastri, 1998; Skinner, 1990).

Some analysts disagree with the notion that SIFCs enhance market efficiency. This contingent believes that index arbitrage programs may actually transmit volatility from the index futures market to the underlying equities market. (5) They cite the stock market crash of October 1987 as an example in which price pressure originating in the index futures market resulted in a collapse in stock prices. In response to these allegations, Miller (1991) argues that SIFC trading only appeared to lead the market because prices of many stocks in the index were not updated from the previous day's close. To date, the impact of SIFC trading on the underlying stock market is an unresolved empirical question.

Although options on the underlying stocks may be used in order to hedge a portfolio, SIFC may be the preferred vehicle for reasons of convenience, low cost, and high liquidity. These characteristics may be especially important for security dealers because portfolio risk is hypothesized to be a major determinant of the dealer's bid/ask spread on a given stock (see Demsetz, 1968; Stoll, 1978a; Tinic, 1972). For example, Stoll (1978a) hypothesizes that both systematic and unsystematic risk should matter to the securities dealer because the marginal position of a dealer at any point in time is undiversified. His empirical findings (Stoll, 1978b) support this hypothesis. In contrast to Stoll's research, Benston and Hagerman (1974) argue that systematic risk should not affect the spread because the dealer's expected return from holding the security compensates him/her for all relevant risks. Their empirical work supports their position. More recently, Tripathy and Peterson (1991) extend empirical testing of the re lationship between risk and the bid/ask spread to account for possible intertemporal changes in systematic risk, and they conclude that systematic risk has no effect on the bid/ask spreads of OTC stocks. We speculate that differences in systematic risk impacts between Stoll (1978b) and Tripathy and Peterson may be due to the fact that Stoll's work predates the advent of SIFC, while Tripathy and Peterson's data follows introduction of SIFC trading.

3. Hedging dealer inventory with stock index futures

The simplest hedging strategy using SIFCs involves purchasing the number of contracts given by:

[Portfolio Value x Portfolio Beta]/[Current Stock Index Value x 500].

Assuming that dealers are more inclined to carry inventory that is easily hedged, we hypothesize that new hedging opportunities brought about by the advent of SIFCs will reduce dealer inventory holding costs. Competition among dealers will induce them to pass at least part of the cost savings onto investors in the form of narrowed bid/ask spreads. Moreover, we believe that even a partial hedge may significantly lower the dealer's inventory holding costs. (6)

As a result of the above considerations, we test empirically the hypothesis that SIFC trading is associated with reduced bid/ask spreads and with decreased sensitivity of bid/ask spreads to changes in portfolio risk. In addition, we explore the possibility that factors may exist that effectively constrain the market maker's use of SIFC. An example of such a factor is the round-trip cost of transacting. Walter and Welles (1988) estimate typical round-trip transactions costs, including internal order clearing, brokerage, and market spread costs, to be approximately US$55. Although this may represent a transactions cost savings relative to alternative hedging techniques, it is still probably unrealistic to expect market makers to hedge indiscriminately across all stocks.

Inspired by Walter and Welles (1988), we examine whether security dealers exercise timing and selectivity in their hedging activities. The terms timing and selectively are traditionally used to describe investment performance criteria (Admati, Bhattacharya, Pfleiderer, & Ross, 1986). In the context of this research, we refer to "timing" as the ability of dealers to hedge inventory risk due to market swings. Walter and Welles (1988) report that specialist firms produce better results when they hedge against fundamental shifts in the market return distribution rather than against day-to-day fluctuations. We refer to "selectivity" as the ability of dealers to pick a subset of stocks to hedge so as to maximize the benefits of hedging. In light of these considerations, we hypothesize that bid/ask spread reductions due to the effect of hedging SIFCs on dealer inventory risk will be most pronounced during periods of high market volatility and for high-beta stocks.

4. Data, methodology, and results

4.1. Sample description

Table 1 lists the dates and exchanges upon which organized trading in four major SIFCs was introduced. We define the SIFC trading introduction period as the 51 trading days between February 24, 1982 and May 6, 1982. This time interval spans the introduction of the first three SIFCs. (7) We further define pre- and post-SWC introduction periods as the 5-year period preceding February 24, 1982, and the 5-year period following May 6, 1982, respectively. We analyze a sample consisting of all 2495 stocks trading on the NASDAQ system for which daily returns and bid/ask spreads are available from the Center for Research in Securities Prices NASDAQ database for a 10-year period surrounding the SIFC introduction period described above.

4.2. Pre- and the post-SIFC introduction period comparison

Table 2 presents descriptive statistics and mean difference t tests for percentage bid/ask spreads, betas, systematic risk measures, and total risk measures for the pre- and post-SIFC introduction periods. Statistical tests reveal that the mean percentage bid/ask spread is smaller following SIFC introduction at conventional significance levels. Moreover, the means of the three risk measures are significantly higher following SIFC introduction. These findings are especially interesting in light of the fact that bid/ask spreads are thought to be positively related to risk (see Stoll, 1978b). We conclude from these findings that observed bid/ask spread reductions following SIFC trading are likely not due to lower intrinsic risk in stock prices.

Next, we partition our sample into quintiles along three dimensions, namely, market value, return volatility, and beta. We then perform a series of mean difference t tests. Table 3, Panel A shows quintile means before and after SIFC introduction, along with the results of t tests by market value quintile. Table 3, Panel B reports analogous statistics by return volatility quintile. Table 3, Panel C reports analogous statistics by beta quintile. We find that mean bid/ask spreads increase (decrease) monotonically with market value (return volatility). Moreover, the collective evidence from Table 3 suggests that (excluding the lowest quintiles for return volatility and beta) the bid/ask spread reduction observed in the post-SIFC introduction period is significant at the 10% level or better in all cases, regardless of how the sample is partitioned. We interpret these findings as evidence that the potential benefits to SIFC hedging span equities of all sizes and risks.

4.3. Bid/ask spreads and portfolio risk

We test whether there exists a positive relationship between observed bid/ask spreads and portfolio risk, and further, whether this relationship is less elastic in the period following SIFC introduction. These tests are based upon the following regression model:

SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i] RIS[K.sub.i] + [[epsilon].sub.i] (1)

where subscript i = index value representing stock I, SPREA[D.sub.i] = [As[k.sub.i] - Bi[d.sub.i]]/[(As[k.sub.i] + Bi[d.sub.i])/2], RIS[K.sub.i] = portfolio risk of holding stock i, SIF[C.sub.i] = 1 if the observation is drawn from the post-SIFC introduction period, and zero otherwise, SIF[C.sub.i].RIS[K.sub.i] = dummy variable for the regression slope associated with RIS[K.sub.i], and [[epsilon].sub.i] = error term, assumed to meet OLS specifications.

We estimate Eq. (1) using a pooled sample drawn from the pre- and post-SIFC trading periods. (8) SPREA[D.sub.i] is a 5-year average value, with the 5-year average computed from data prior to (after) SIFC introduction if the observation is drawn from the pre- (post-) SIFC introduction period. Two alternative measures are computed to represent RIS[K.sub.i]. First, we define RIS[K.sub.i] as [[beta].sub.i.sup.2] x [[sigma].sub.m.sup.2], where [[beta].sub.i] is the stock's OLS beta estimated from a market model regression, and [[sigma].sub.m.sup.2] is the variance of NASDAQ market returns. We define RIS[K.sub.i] alternatively as the stock return variance, [[sigma].sub.i.sup.2]. We estimate regressions using the second RIS[K.sub.i] definition in light of well-known problems in estimating beta (i.e., infrequent trading, nonsynchronous stock and index returns, and market model nonstationarities). Moreover, Stoll (1978b) posits that both systematic risk and total risk affect the bid/ask spread.

SIF[C.sub.i] and SIF[C.sub.i]*RIS[K.sub.i] are intercept and slope dummy variables designed to test for structural changes in market microstructure associated with the introduction of SIFCs. We predict that the coefficient associated with SIF[C.sub.i] will be significantly negative because we believe that the advent of SIFCs should be associated with smaller bid/ask spreads for reasons outlined earlier. Furthermore, we predict that the slope dummy variable SIF[C.sub.i]*RIS[K.sub.i] will be significantly negative because we believe that SIFC trading to some degree immunizes dealers from the effects of portfolio risk, thereby reducing the elasticity of the spread with respect to risk.

We report full-sample results for estimation of Eq. (1) in Table 4. We report quintile results for regressions estimated from subsamples disaggregated by market value, return volatility, and beta in Table 5, Panels A through C, respectively. For the sake of brevity, we report regression results only for RIS[K.sub.i] proxied by total risk in Table 5. Based upon our interpretation of the coefficients and significance tests reported in Tables 4 and 5, we present three salient findings.

First, we observe from Tables 4 and 5 that, in general, portfolio risk is positively related to the bid/ask spread. This implies that security dealers may be affected by portfolio risk, and thus there exists the potential for dealers to benefit from hedging opportunities. Second, based upon our interpretation of the significant negative coefficients associated with the slope dummy variable SIFC.RISK reported in Tables 4 and 5, Panel C, the sensitivity of bid/ask spreads with respect to portfolio risk is reduced in the post-SIFC trading period. We interpret this finding as evidence that security dealers reduced their required compensation for market-related risk following the introduction of SIFCs. Table 5, Panels A and B report mixed results. Finally, the significantly negative coefficients for variable SIFC reported in Tables 4 and 5, Panel C suggest that there is a secular decline in the percentage bid/ask spread in the post-SIFC period. We speculate that this decline may have resulted from increased tradin g volume (and liquidity), and/or from reduced adverse selection costs in the NASDAQ market following SIFC introduction. Again, we observed mixed results for variable SIFC in Table 5, Panels A and B. (9)

We would like to include market microstructure variables, such as trading volume and stock price, in our estimated regressions. Unfortunately, trading volume data are not available for stocks in the pre-SIFC trading period. Nevertheless, we attempt to incorporate volume effects in our models by dividing the sample into volume quintiles using the available post-SIFC volume data. We then estimate regression models by quintile. We anticipate that volume (by quintile) will interact with the relationship between the bid/ask spread and the explanatory variables. We also include the stock price, with the expectation that it is negatively correlated with the percentage bid/ask spread. Finally, as an alternative to the stock price, we include market capitalization in the regression specification. We present the results from these attempts to incorporate general microstructure variables in the regressions in Table 6.

Results from the experiments with market microstructure variables yield three observations. First, as anticipated, stock prices and, alternatively, market capitalization values are significantly negatively related to the percentage bid/ask spread, with the exception of market capitalization in the case of the highest volume quintile. Second, the inclusion of the microstructure variables improves the overall significance and explanatory power of the estimated regressions. Finally, the slope and intercept dummy coefficients are uniformly significantly negative through the middle three quartiles in Table 6, Panels A and B.

Overall, our regression models reveal three findings. First, percentage bid/ask spreads are lower in the post-SIFC period. Second, while the percentage bid/ask spread is negatively related to portfolio risk generally, the percentage spread is less sensitive to changes in portfolio risk in the post-SIFC period. Finally, the observed relationships are robust with respect to controls for market capitalization, return volatility, beta, stock price, and trading volume.

4.4. Timing and selectivity effects

Our final set of tests is designed to determine whether the magnitude of bid/ask spread changes between pre- and post-SIFC trading periods is greater during periods of high market volatility and for high beta stocks, as discussed in Section 3. For the analysis with respect to timing, we select four quarters from our pre- and post-SIFC intervals representing four regimes: (1) high volatility, pre-SIFC, (2) low volatility, pre-SIFC, (3) high volatility, post-SIFC, and (4) low volatility, post-SIFC. We then estimate two regression models, one corresponding to low volatility markets and the other corresponding to high volatility markets. For the analysis with respect to selectivity, we estimate two regression models, one for below median beta stocks and the other for above median beta stocks.

In order to implement the timing tests, we use the CRSP NASDAQ database to construct a series of rolling 30-day return variances for returns on the NASDAQ index. The quarterly averages from this rolling series are illustrated in Fig. 1 for the pre-SIFC period and in Fig. 2 for the post-SIFC period. For the pre-SIFC period, 1980 II and 1979 III are the quarters associated with the highest and lowest index return variances, respectively. For the post-SIFC period, the corresponding maximum and minimum average index variance quarters are 1982 IV and 1985 II, respectively. (10)

Table 7, Panel A presents regression estimates for Eq. (1) for data from pre- and post-SIFC high variance quarters, and also for data from pre- and post-SIFC low variance quarters. All regression coefficients reported in Table 7, Panel A are significant at the 10% level or better, except for the slope dummy variable in the low variance model. All other estimated coefficients are similar to our original results. We interpret these results as evidence that hedging opportunities afforded by SIFC trading may be more valuable to dealers and investors during periods of high market volatility than is the case during periods of low market volatility. (11)

Table 7, Panel B presents regression estimates for Eq. (1) for data from pre- and post-SIFC high beta stocks, and also for data from pre- and post-SIFC low beta stocks. As was the case in Table 7, Panel A, all regression coefficients reported in Table 7, Panel B are significant at the 10% level or better, except for the slope dummy variable in the low beta model. All other estimated coefficients are similar to our original results. We interpret these results as evidence that the hedging opportunities afforded by SIFC trading may be most valuable for hedging higher beta stocks.

Before concluding the current set of tests, we conduct an alternative test of the selectivity hypothesis. We estimate the following regression (Eq. (2)) in order to determine whether high beta stocks experienced a larger decline in bid/ask spreads than low beta stocks:

CHSPREAD = -0.02[6.sub.(-7.5)] - 0.008[2.sub.(-2.6)] BETA (2)

where CHSPREAD = the difference between 5-year average bid/ask spreads for each stock estimated over the pre- and post-SIFC trading period, BETA = the OLS estimate of beta for the same stock estimated using 5 years of monthly returns over the pre-SIFC trading period. We interpret the statistically significant negative slope coefficient as evidence to support the hypothesis that higher beta stocks exhibit a larger decline in bid/ask spreads in the post-SIFC trading period. Overall, the results from our timing and selectivity tests are consistent with the notion that dealers and market makers either choose, or are able, to more effectively hedge inventory risk with SIFCs during periods of high market volatility and for high beta stocks.

5. Conclusion

We analyze the behavior of bid/ask spreads on NASDAQ stocks before and after the introduction of stock index futures trading. Our analysis reveals the following findings. First, we find that a stock's bid/ask spread is positively related to its systematic risk. We interpret this finding as evidence that dealers "price" portfolio risk. Dealers may therefore benefit from hedging this risk by trading SIFCs. Moreover, we find that the sensitivity of spreads to changes in portfolio risk is reduced in the post-SIFC trading period. We interpret this finding as evidence that the introduction of SIFC trading has allowed security dealers to reduce their portfolio risk.

As a result, dealers are able to reduce bid/ask spreads in the market. Second, we document an overall decline in percentage bid/ask spreads in the post-SIFC trading period, even after controlling for possible structural changes caused by SIFC trading. We attribute this reduction in spreads to the increased availability of cost-effective hedging opportunities provided by SIFCs. (12)

Since hedging costs represent a tangible expense that must be borne by market participants desiring to hedge their risks, we also examine whether dealers may benefit from timing their hedging activities, and/or selectively employing hedging strategies primarily for high systematic risk stocks. While we find that the positive relationship between bid/ask spreads and portfolio risk is lessened following SIFC introduction, we find that this phenomenon is more prominent during periods of high market volatility. We also find that higher beta stocks exhibit a larger reduction in bid/ask spreads in the post-SIFC trading period than lower beta stocks. We interpret these findings as evidence consistent with the notion that rational utility-maximizing traders concentrate their hedging efforts and resources on higher risk stocks and during periods of greater market volatility.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

Table 1 Listing dates of the major SIFCs Stock index Exchange Value Line Composite Index Kansas City Board of Trade S&P 500 Index Chicago Mercantile Exchange New York Stock Exchange Composite Index New York Futures Exchange Major Market Index Chicago Board of Trade Stock index Date listed Value Line Composite Index February 24, 1982 S&P 500 Index April 21, 1982 New York Stock Exchange Composite Index May 6, 1982 Major Market Index July 23, 1984 Table 2 Summary statistics from the pre- and post-SIFC trading periods Pre-index period (a) Post-index period (a) Variables Mean S.D. Mean Percentage bid/ask spreads 10.61% 8.76% 6.75% Beta 0.989 0.547 1.040 Systematic risk 0.003 0.004 0.004 Total risk 0.023 0.032 0.028 Sample size 1248 1247 Post-index period (a) Variables S.D. Two Sample t test (b) Percentage bid/ask spreads 6.40% - 12.56 *** Beta 0.808 1.85 * Systematic risk 0.001 8.56 *** Total risk 0.095 1.76 * Sample size The percentage bid/ask spread of stock i, measured as (Ask - Bid)/(Ask + Bid)0.5. Beta is the OLS beta estimated from the market model based on the NASDAQ index. Systematic risk is measured as [B.sub.i.sup.2][[sigma].sup.2]([R.sub.m]) where B is the beta estimated from the market model and [[sigma].sup.2]([R.sub.m]) is the variance of the market returns (NASDAQ index). Total risk is the return variance of stock i. (a) The pre- and post-SIFC trading periods are defined, respectively, as the 5-year period preceding and following the 51 trading days between February 24, 1982, and May 6, 1982 when the first three of the four SIFC trading were introduced. All stocks trading in the NASDAQ system are evenly split and randomly assigned either to the pre- or post-SIFC period. (b) Based on a comparison of mean values from the pre- and post-index periods. * Indicates significance at the .10 level. ** Indicates significance at the .05 level. *** Indicates significance at the .01 level. Table 3 Comparison of percentage bid/ask spread between the pre- and post-index period based on market value, return volatility, and beta quintiles Panel A: Comparison of spread between the pre- and post-index period by market value quintiles Quintile I: Quintile II: median market median market value = US$3,506,130 value = US$9,867,010 Pre- Post- Pre- index index Two index period: period: sample period: mean/ mean/ t statistic mean/ S.D. S.D. (P value) S.D. Bid/ask 23.05/ 17.27/ 2.90 11.67/ spread (%) 13.96 12.36 (.0043) 4.94 Quintile II: Quintile III: median market median market value = US$9,867,010 value = US$21,758,060 Post- Pre- index Two index period: sample period: mean/ t statistic mean/ S.D. (P value) S.D. Bid/ask 6.83/ 7.55 8.86/ spread (%) 3.44 (.0001) 5.51 Quintile III: Quintile IV: median market median market value = US$21,758,060 value = US$48,117,680 Pre- index Two index index period: sample period: period: mean/ t statistic mean/ mean/ S.D. (P value) S.D. S.D. Bid/ask 4.68/ 6.30 5.89/ 2.55/ spread (%) 2.91 (.0001) 3.26 9.11 Quintile IV: Quintile V: median market median market value = value = US$211,722,070 US$48,117,680 Pre- Two index index sample period: period: t statistic mean/ mean/ (P value) S.D. S.D. Bid/ask 3.23 3.04/ 2.41/ spread (%) (.0015) 1.38 1.97 Quintile V: median market value = US$211,722,07 0 Two sample t statistic (P value) Bid/ask 2.43 spread (%) (.0160) Panel B: Comparison of spread between the pre- and post-index period by return volatility quintiles Quintile I: Quintile II: median volatility median volatility level = 0.00014 level = 0.00033 Pre- Post- Pre- index index Two index period: period: sample period: mean/ mean/ t statistic mean/ S.D. S.D. (P value) S.D. Bid/ask 5.84/ 4.71/ 1.41 6.17/ spread (%) 4.26 6.06 (.1592) 5.19 Quintile II: Quintile III: median volatility median volatility level = 0.00033 level = 0.00067 Post- Pre- index Two index index period: sample period: period: mean/ t statistic mean/ mean/ S.D. (P value) S.D. S.D. Bid/ask 3.70/ 2.02 9.11/ 7.05/ spread (%) 10.21 (.0451) 6.88 7.43 Quintile III: Quintile IV: median median volatility volatility level = value = 0.00107 0.00067 Pre- Two index index Two sample period: period: sample t statistic mean/ mean/ t statistic (P value) S.D. S.D. (P value) Bid/ask 1.90 13.37/ 7.74/ 3.45 spread (%) (.0586) 11.48 10.17 (.0007) Quintile V: median volatility value = 0.00228 Pre- index index Two period: period: sample mean/ mean/ t statistic S.D. S.D. (P value) Bid/ask 18.00/ 10.49/ 4.42 spread (%) 13.02 9.10 (.0001) Panel C: Comparison of spread between the pre- and post-index period by beta quintiles Quintile I: Quintile II: median beta median beta level = 0.0657 level = 0.2230 Pre- Post- Pre- Post- index index Two index index period: period: sample period: period: mean/ mean/ t statistic mean/ mean/ S.D. S.D. (P value) S.D. S.D. Bid/ask 12.39/ 10.84/ 0.71 11.53/ 8.15/ spread (%) 13.68 15.23 (.4800) 10.41 8.54 Quintile II: Quintile III: median beta median beta level = level = 0.4078 0.2230 Pre- Two index index Two sample period: period: sample t statistic mean/ mean/ t statistic (P value) S.D. S.D. (P value) Bid/ask 2.35 10.67/ 4.84/ 5.03 spread (%) (.0197) 9.45 5.40 (0.0001) Quintile IV: Quintile V: median beta median beta level = 0.6339 Level = 1.1454 Pre- Pre- index index Two index period: period: sample period: mean/ mean/ t statistic mean/ S.D. S.D. (P value) S.D. Bid/ask 9.48/ 4.97/ 4.20 8.41/ spread (%) 7.89 6.26 (.0001) 6.68 Quintile V: median beta Level = 1.1454 index Two period: sample mean/ t statistic S.D. (P value) Bid/ask 4.90/ 4.44 spread (%) 3.22 (.0001) Table 4 Change in the relation between dealer's spreads and portfolio risk following the introduction of SIFC trading RISK = systematic risk RISK = total risk [[delta].sub.0] 0.10 (35.3 ***) 0.08 (32.0 ***) [[delta].sub.1] 1.09 (2.2 **) 1.02 (16.4 ***) [[delta].sub.2] - 0.04 (- 9.5 ***) - 0.02 (- 5.2 ***) [[delta].sub.3] - 0.89 (- 1.8 *) - 0.89 (- 13.3 ***) Sample size 2417 2417 Adj. [R.sup.2] .061 .159 F 52.91 *** 53.50 *** The following regression is used: SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i]*RIS[K.sub.i]. t values are reported in parenthesis. SPREA[D.sub.i] is the percentage bid/ask spread of stock i, measured as (Ask - Bid)/(Ask + Bid)0.5. RISK is the dealer's portfolio risk. Two alternative measures are used to represents this risk: systematic risk is measured as [B.sub.i.sup.2][[sigma].sup.2]([R.sub.m]) where B is the stock's OLS beta estimated from the market model and [[sigma].sup.2]([R.sub.m]) is the variance of the market returns (NASDAQ index); total risk, [[sigma].sup.2]([R.sub.i]), is the return variance of stock i. SIF[C.sub.i] is a dummy variable that assumes a value of 0 if stock i is drawn from the pre-index trading period, and 1 if it is drawn from the post-trading period. SIF[C.sub.i]*RIS[K.sub.i] is the cross product of SIFC and RISK. The pre- and post-index trading periods are defined, respectively, as the 5-year period preceding and following the 51 trading days between February 24, 1982, and May 6, 1982, when the first three of the four SIFC trading were introduced. t values are reported in parenthesis. * Indicates significance at the .10 level. ** Indicates significance at the .05 level. *** Indicates significance at the .01 level Table 5 Estimation of the spread-regression based on market value, return volatility, and beta quintiles Panel A: Spread regression coefficients on samples based on market value quintiles, and the associated t and P values Quintile I Quintile II Intercept: ([[delta].sub.0])/t/P) .53/9.2/.00 .24/8.9/.00 RISK ([[delta].sub.1])/t/P) .05/5.6/.00 .02/5.0/.00 SIFC ([[delta].sub.2])/t/P) - .12/ - 1.2/.23 - .09/ - 2.0/.04 SIFC*RISK ([[delta].sub.3])/t/P) - .02/ - 1.24/.22 - .01/ - 1.7/.09 F value/adj. [R.sup.2] 12.64/.09 10.60/.08 Quintile III Quintile IV Intercept: ([[delta].sub.0])/t/P) .22/8.5/.00 .09/3.1/.00 RISK ([[delta].sub.1])/t/P) .02/6.1/.00 .01/1.5/.13 SIFC ([[delta].sub.2])/t/P) - .06/ - 1.5/.14 .04/0.8/.40 SIFC*RISK ([[delta].sub.3])/t/P) - .01/ - 1.0/.30 .01/1.2/.22 F value/adj. [R.sup.2] 19.09/.14 5.06/.34 Quintile V Intercept: ([[delta].sub.0])/t/P) .05/3.8/.00 RISK ([[delta].sub.1])/t/P) .002/1.5/.13 SIFC ([[delta].sub.2])/t/P) .03/2.0/.05 SIFC*RISK ([[delta].sub.3])/t/P) .001/2.4/.02 F value/adj. [R.sup.2] 14.7/.11 Panel B: Spread regression coefficients on samples based on return volatility quintiles, and the associated t and P values Quintile I Quintile II Intercept: ([[delta].sub.0])/t/P) .04/0.5/.59 .28/1.4/.16 RISK ([[delta].sub.1])/t/P) - .002/ - 0.2/.85 .03/1.1/.28 SIFC ([[delta].sub.2])/t/P) - .19/ - 1.7/.10 - .11/ - 0.4/.68 SIFC*RISK ([[delta].sub.3])/t/P) - .02/ - 1.6/.11 - .01/ - 0.3/.76 F value/adj. [R.sup.2] 2.25/.01 4.92/.03 Quintile III Quintile IV Intercept: ([[delta].sub.0])/t/P) .25/0.97/.33 .65/2.1/.04 RISK ([[delta].sub.1])/t/P) .02/0.6/.53 .08/1.7/.10 SIFC ([[delta].sub.2])/t/P) - .01/ - 0.04/.96 .17/0.4/.72 SIFC*RISK ([[delta].sub.3])/t/P) .001/0.03/.98 .03/0.5/.63 F value/adj. [R.sup.2] 4.15/.03 10.50/.08 Quintile V Intercept: ([[delta].sub.0])/t/P) .71/5.8/.00 RISK ([[delta].sub.1])/t/P) .09/4.4/.00 SIFC ([[delta].sub.2])/t/P) - .19/ - 1.1/.28 SIFC*RISK ([[delta].sub.3])/t/P) - .02/ - 0.8/.43 F value/adj. [R.sup.2] 15.77/.11 Panel C: Spread regression coefficient on samples based on beta quintiles, and the associated t and P values Quintile I Quintile II Intercept: ([[delta].sub.0])/t/P) .70/11.9/.00 .74/19.6/.00 RISK ([[delta].sub.1])/t/P) .07/9.7/.00 .08/12.6/.00 SIFC ([[delta].sub.2])/t/P) - .19/ - 2.2/.03 - .29/ - 4.1/.00 SIFC*RISK ([[delta].sub.3])/t/P) - .02/ - 1.7/.09 - .03/ - 3.3/.00 F value/adj. [R.sup.2] 44.98/.27 77.45/.40 Quintile III Quintile IV Intercept: ([[delta].sub.0])/t/P) .70/21.1/.00 .59/14.8/.00 RISK ([[delta].sub.1])/t/P) .08/18.0/.00 .07/12.5/.00 SIFC ([[delta].sub.2])/t/P) - .33/ - 7.1/.00 - .22/ - 4.1/.00 SIFC*RISK ([[delta].sub.3])/t/P) - .04/ - 5.7/.00 - .02/ - 3.2/.00 F value/adj. [R.sup.2] 100.70/.59 95.40/.45 Quintile V Intercept: ([[delta].sub.0])/t/P) .46/15.7/.00 RISK ([[delta].sub.1])/t/P) .06/13.0/.00 SIFC ([[delta].sub.2])/t/P) - .22/ - 4.5/.00 SIFC*RISK ([[delta].sub.3])/t/P) - .03/ - 3.8/.00 F value/adj. [R.sup.2] 85.70/.42 The following regression is used: SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i]*RIS[K.sub.i]. SPREA[D.sub.i] is the percentage bid/ask spread of stock i. RIS[K.sub.i] is portfolio risk measured as the return variance of stock i, [[sigma].sup.2]([R.sub.i]). SIF[C.sub.i] is a dummy variable that assumes a value of 0 if stock i is drawn from the pre-index trading period, and 1 if it is drawn from the post-index trading period. SIF[C.sub.i]*RIS[K.sub.i] is the cross product of SIFC and RISK. Table 6 Estimation of the spread-regressions based on trading volume quintiles Panel A: The following regression is used: SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i]*RIS[K.sub.i] + [[delta].sub.4]PRIC[E.sub.i]. SPREA[D.sub.i] is the percentage bid/ask spread of stock i. RISK is portfolio risk measured as the return variance of stock i, [[sigma].sup.2]([R.sub.i]). RIS[K.sub.i] is a dummy variable that assumes a value of 0 if stock i is drawn from the pre-idex trading period, and I if it is drawn from the post-index trading period. SIF[C.sub.i]*RIS[K.sub.i] is the interaction term between SIFC and RISK. PRICE is the market price of stock i. Quintile I Quintile II Intercept: ([[delta].sub.0]/t/P) .17/3.1/.00 .32/12.3/.00 RISK ([[delta].sub.1]/t/P) .00009/0.01/.99 .03/7.7/.00 SIFC ([[delta].sub.2]/t/P)) -.01/-0.1/.95 -.16/-4.8/.00 SIFC.RISK ([[delta].sub.3]/t/P) .002/0.16/.87 -.02/-4.1/.00 PRICE ([[delta].sub.3]/t/P) -.05/-7.3/.00 -.02/-4.9/.00 F value/adj. [R.sup.2] 27.11/.26 72.73/.37 Quintile III Quintile IV Intercept: ([[delta].sub.0]/t/P) .23/14.4/.00 .18/25.8/.00 RISK ([[delta].sub.1]/t/P) .02/7.7/.00 .02/14.9/.00 SIFC ([[delta].sub.2]/t/P)) -.08/-3.9/.00 -.05/-5.5/.00 SIFC.RISK ([[delta].sub.3]/t/P) -.01/-3.4/.00 -.01/-5.1/.00 PRICE ([[delta].sub.3]/t/P) -.02/-8.0/.00 -.01/-11.5/.00 F value/adj. [R.sup.2] 129.41/.43 266.08/.58 Quintile V Intercept: ([[delta].sub.0]/t/P) .12/44.0/.00 RISK ([[delta].sub.1]/t/P) .01/29.2/.00 SIFC ([[delta].sub.2]/t/P)) -.003/-0.7/.50 SIFC.RISK ([[delta].sub.3]/t/P) -.0004/-0.9/.39 PRICE ([[delta].sub.3]/t/P) -.003/-10.6/.00 F value/adj. [R.sup.2] 643.63/.73 Panel B: The following regression is used: SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i]*RIS[K.sub.i] + [[delta].sub.4]MVALU[E.sub.i]. SPREA[D.sub.i] is the percentage bid/ask spread of stock i. RISK is portfolio risk measured as the return variance of stock i, [[sigma].sup.2]([R.sub.i]). RIS[K.sub.i] is a dummy variable that assumes a value of 0 if stock i is drawn from the pre-idex trading period, and I if it is drawn from the post-index trading period. SIF[C.sub.i]*RIS[K.sub.i] is the interaction term between SIFC and RISK. MVALUE is the market capitalization of stock i. Quintile I Quintile II Intercept: ([[delta].sub.0]/t/P) .51/8.6/.00 .43/16.6/.00 RISK ([[delta.sub.1]/t/P) .03/3.1/.00 .04/10.6/.00 SIFC ([[delta].sub.2]/t/P) .07/0.7/.46 -.14/-4.3/.00 SIFC.RISK ([[delta].sub.3]/t/P) .01/0.9/.39 -.02/-3.7/.00 MVALUE ([[delta].sub.3/t/P) -.03/-4.6/.00 -.01/-5.4/.00 F value/adj. [R.sup.2] 17.68/.18 74.83/.37 Quintile III Quintile IV Intercept: ([[delta].sub.0]/t/P) .35/21.5/.00 .22/27.7/.00 RISK ([[delta.sub.1]/t/P) .03/11.0/.00 .02/17.4/.00 SIFC ([[delta].sub.2]/t/P) -.08/-3.6/.00 -.04/-3.7/.00 SIFC.RISK ([[delta].sub.3]/t/P) .01/-3.4/.00 -.01/-3.6/.00 MVALUE ([[delta].sub.3/t/P) -.01/-9.0/.00 -.004/-7.7/.00 F value/adj. [R.sup.2] 136.27/.44 228.57/.54 Quintile V Intercept: ([[delta].sub.0]/t/P) .12/41.2/.00 RISK ([[delta.sub.1]/t/P) .01/31.5/.00 SIFC ([[delta].sub.2]/t/P) -.004/-1.0/.31 SIFC.RISK ([[delta].sub.3]/t/P) -.005/-1.0/.33 MVALUE ([[delta].sub.3/t/P) .0004/2.29/.02 F value/adj. [R.sup.2] 554.66/.70 Two sets of spread regression are used that estimate the change in the percentage bid/ask spreads following the introduction of SIFC, after controlling for the effects of changes in portfolio risk associated with the stocks, its price level or market capitalization. Both regressions are estimated over subsamples that ae formed by subdividing the full sample into quintiles based on average trading volume. Since complete trading volume data is available only for the post-index trading period, the quintiles are based on the average trading volume for each stock in the post-index trading interval. The quintiles are in ascending order of the average trading volume. Table 7 Evidence on the market maker's exercise of timing and selectivity in the use of SIFC Panel A: Change in the Spread-Systematic risk relation during periods of low and high market volatility chosen from the pre- and post-index trading periods. [[delta].sub.0] [[delta].sub.1] Low market volatility 0.09 (27.4) 1.49 (2.7) period (a) High market volatility 0.10 (24.5) 2.49 (2.9) period (b) [[delta].sub.2] [[delta].sub.3] Low market volatility -0.04 (-7.7) -0.71 (-0.9) * period (a) High market volatility -0.03 (-4.6) -4.45 (-3.9) period (b) F ratio Adj. [R.sup.2] Low market volatility 40.91 .62 period (a) High market volatility 37.53 .60 period (b) Panel B: Change in the Spread -- Systematic risk relation around the introduction of SIFC trading for stocks with high and low systematic risk. [[delta].sub.0] [[delta].sub.1] Low systematic risk stocks (c) 0.13 (29.5) 8.57 (6.7) High systematic risk 0.06 (20.8) 3.71 (9.9) stocks (d) [[delta].sub.2] [[delta].sub.3] Low systematic risk stocks (c) -0.06 (-10.7) -1.08 (-0.6) * High systematic risk -0.01 (-2.7) -3.47 (-9.0) stocks (d) F ratio Adj. [R.sup.2] Low systematic risk stocks (c) 91.23 .18 High systematic risk 65.50 .14 stocks (d) The following regression is used: SPREA[D.sub.i] = [[delta].sub.0] + [[delta].sub.1]RIS[K.sub.i] + [[delta].sub.2]SIF[C.sub.i] + [[delta].sub.3]SIF[C.sub.i]*RIS[K.sub.i]. t values are reported in parenthesis. SPREA[D.sub.i] is the percentage bid/ask spread of stock i, measured as (Ask - Bid)/(Ask + Bid)0.5. RISK is portfolio risk. Two alternative measures are used to represents this risk: Systematic risk is measured as [B.sub.i.sup.2][[sigma].sup.2]([R.sub.m]), where B is the stock's OLS beta estimated from the market model and [[sigma].sup.2]([R.sub.m]) is the variance of the market returns (NASDAQ index); Total risk, [[sigma].sup.2]([R.sub.i]), is the return variance of stock i. SIF[C.sub.i] is a dummy variable that assumes a value of 0 if stock i is drawn from the pre-SIFC trading period, and 1 if it is drawn from the post-trading period. SIF[C.sub.i]*RIS[K.sub.i] is the cross product of SIFC and RISK. (a) Pooled samples drawn from the third quarter of 1979 in the pre-index futures period and the fourth quarter of 1985 in the post-index futures period. (b) Pooled samples drawn from the first quarter of 1980 in the pre-index futures period and the fourth quarter of 1982 in the post-index futures period. (c) Stocks in this sample have below median systematic risk. (d) Stocks in this sample have above median systematic risk. * Indicates not significant at the 10% level. All other coefficients are significant at that level.

Acknowledgments

We would like to thank Peggy Swanson, J. David Diltz, two anonymous reviewers, and the editor of this journal for their constructive suggestions. Any remaining errors are our own.

(1.) Tel.: +1-940-565-3045.

(2.) To the extent that the borrowing and lending rates are applicable to large market participants who set option prices on the margin, these rates can be favorable to the typical options trader.

(3.) Under the current rules, a short sale can be executed only when the previous nonzero price change is positive. Also, use of options market avoids loss of frill proceed of short-sale transaction in the stock market.

(4.) Stephan and Whaley (1990) report that price adjustments across the stock and options markets are accomplished in 15 minutes. In the case of block trades, Kumar, Sarin, and Shastri (1992) observe that significant lead--lag relationship of returns between the two occur within only a 30-minute interval.

(5.) See Miller (1991) for a detailed discussion of this issue.

(6.) As Figlewski and Kon (1982) observe, only a stock portfolio whose composition is identical to the index can be perfectly hedged. Stock index futures, which are essentially cross-hedges, will therefore provide only incomplete hedges against systematic risk in the marker's portfolio.

(7.) We do not consider the incremental effect of the fourth SIFC that was introduced July 23, 1984. We exclude the fourth SIFC from our consideration because we believe that our current 4-month introduction period would introduce smaller error in our statistical estimates than 30-month window that would be required if we were to include the fourth SIFC in our testing. Moreover, we trade the other three contracts prior to July 23, 1984.

(8.) We recognize that this regression specification is incomplete in the sense that it does not account for other bid/ask spread determinants, such as volume, number of dealers, and stock price. We incorporate some of these variables in subsequent testing.

(9.) In order to determine the stability of cross-sectional regression coefficients, we repeated our regression estimation using two and four intervals instead of the year interval reported in the body of the paper. We did not find any qualitative difference in the sign, magnitude, or regression coefficients estimated from these alternate windows.

(10.) We observe that the market variance levels in the selected high and low quarters from the pre-index periods are noticeably higher than the corresponding quarters in the post-index period. To preserve our results from contamination by this bias, we select quarters with approximately equal high and low levels of market variance from both periods:

Pre-index period Post-index period High variance level 0.0000454 (1,1980) 0.0000431 (IV, 1982) Low variance level 0.0000102 (III, 1979) 0.0000105 (IV, 1985)

(11.) We must note, however, that if market volatility interacts systematically with the spread determinants in our sample, then the results presented in Table 7, Panel A must be regarded as inconclusive.

(12.) We acknowledge, however, that observed changes in bid/ask spreads are also consistent wit

References

Admati, A. R., Bhattacharya, S., Pfleiderer, P., & Ross, S. A. (1986). On timing and selectivity. Journal of Finance, 41, 715-729.

Benston, G. G., & Hagerman, R. L. (1974). Determinants of bid--ask spreads in the over-the-counter market. Journal of Financial Economics, 2, 353-364.

Black, F. (1975). Facts and fantasy in the use of options. Financial Analysts Journal, 31, 61-72.

Conrad, J. (1989). The price effect of options introduction. Journal of Finance, 44, 487-498.

Coughenour, J., & Shastri, K. (1999). Symposium on market microstructure: a review of empirical literature. Financial Review, 34, 1-28.

Cox, J. C., & Rubinstein, M. (1985). Options markets. Englewood Cliffs, NJ: Prentice Hall.

Demsetz, H. (1968). The cost of transacting. Quarterly Journal of Economics, 32, 33-53.

Ederington, L. (1979). The hedging performance of the new futures market. Journal of Finance, 34, 157-170.

Figlewski, S. (1985). Hedging with stock index futures: theory and application in a new market. Journal of Futures Market, 5, 183-199.

Figlewski, S., & Kon, S. J. (1982). Portfolio management with stock index futures. Financial Analysts Journal, 38, 52-60.

Junkus, J. C., & Lee, C. F. (1985). Use of three stock index futures in hedging decisions. Journal of Futures Market, 5, 201-222.

Kumar, R., Sarin, A., & Shastri, K. (1992). The behaviour of option price around large block transactions in the underlying security. Journal of Finance, 47, 879-890.

Kumar, R., Sarin, A., & Shastri, K. (1998). The impact of options trading in the market quality of the underlying security: an empirical analysis. Journal of Finance, 53, 717-732.

Miller, M. H. (1991). Financial innovations and market volatility. Cambridge, MA: Blackwell Publishers.

Nordhauser, F. (1984). Using stock index futures to reduce market risk. Journal of Portfolio Management, 10, 56-62.

Skinner, D. J. (1990). Options markets and the information content of accounting earnings releases. Journal of Accounting and Economics, 13, 191-211.

Stephan, J., & Whaley, R. (1990). Intraday price change and trading volume relations in the stock and stock options markets. Journal of Finance, 55, 191-220.

Stoll, H. (1978a). The supply of dealer service to the securities markets. Journal of Finance, 33, 1133-1151.

Stoll, H. (1978b). The pricing of security dealer's service: an empirical study of NASDAQ stocks. Journal of Finance, 33, 1153-1172.

Tinic, S. M. (1972). The economics of liquidity services. Quarterly Journal of Economics, 36, 79-93.

Tripathy, N., & Peterson, R. L. (1991). Risk adjustment costs and securities dealer's bid-ask spreads. Journal of Financial Service Research, 5, 131-142.

Walter, R. C., & Welles, E. C. (1988). Hedging the market risk in the specialist portfolio. Presented at the Financial Management Association Annual Meeting, New Orleans.

Weiner, N. S. (1981). The hedging rationale for a stock index futures contract. Journal of Futures Market, 1,59-76.

Salil K. Sarkar (a) *, Niranjan Tripathy (b,1)

(a) Finance and Real Estate Departmnet, University of Texas at Arlington, Box 19499, Arlington, TX 76019, USA

(b) Departmnet of Finanace, Insurance, Real Estate and Law, University of North Texas, Denton, TX 76203, USA

* Corresponding author. Tel.: +1-817-272-3836; fax: +1-817-272-2252.

E-mail addresses: sarkar@uta.edu (S.K. Sarkar), tripathy@unt.edu (N. Tripathy).

Printer friendly Cite/link Email Feedback | |

Author: | Sarkar, Salil K.; Tripathy, Niranjan |
---|---|

Publication: | Review of Financial Economics |

Geographic Code: | 1USA |

Date: | Jan 1, 2002 |

Words: | 8563 |

Previous Article: | CEO compensation, option incentives, and information disclosure. |

Next Article: | A model of broker's trading, with applications to order flow internalization **. |

Topics: |