Trading Analysis of
October 27 and 28, 1997

A Report by the Division of Market Regulation
U.S. Securities and Exchange Commission

September 1998

Although the Commission has authorized publication of this report, it has expressed no view regarding the analyses, findings or conclusions therein.

Note: Graphs and tables appearing in attachments A through D are not available in this electronic version of the report.

Executive Summary

Immediately following October 27 and 28, Chairman Levitt requested that the Division of Market Regulation ("Division") reconstruct the days' events. The findings from this reconstruction are discussed in Part II of this report. The Division was also asked to determine the effects of circuit breakers on market price movements. The Commission's Office of Economic Analysis ("OEA") helped analyze the effects of the circuit breakers on the velocity of market price movements and the quality of market-making on October 27. The determinations regarding circuit breakers are discussed in Part III of this report. Finally, the Division analyzed the operational performance of the markets' automated trading, price-reporting, and clearance and settlement systems on October 27 and 28. In conjunction with the Division's analysis, the Commission's Office of Compliance Inspections and Examinations ("OCIE") inspected the systems performance of major full-service and online broker-dealers. The findings from these reviews are discussed in Part IV of this report.

Overall, the Division's trading reconstruction indicates that the steep market declines on October 27 and the morning of October 28 appear to have been a price correction along the lines of the 6.91% decline on October 13, 1989. On October 27, 1997, the selloff appears to have been prompted by concerns over the potential impact on U.S. corporate earnings of the growing market turmoil in Asia and the repercussions from the potential economic slowdown and deflationary pressures. The Asian market turmoil evidently caused a number of institutional and professional traders to attempt to reduce their equity exposure or increase their hedges in the U.S. markets, either directly through stock sales or indirectly through trades in futures. While various trading strategies and overall liquidity constraints in the stock and futures markets, as well as concerns over the potential early closure of the markets on October 27, may have contributed to the velocity of the price declines, no single factor emerged as the cause for the price volatility during this period. When the selloff reduced U.S. stock prices to attractive levels on the morning of October 28, broad-based buying developed to support a strong, albeit partial, bounce-back in share prices on then-record volumes.

The market decline on October 27 was not of a magnitude to offer a true test of how circuit breakers might operate during more severe declines. Nevertheless, the Division believes that the events on October 27 offered several critical insights into the need to change the circuit breaker procedures then in effect. The Division discussed these issues with the securities markets in the context of their revisions to the circuit breaker procedures which went into effect on April 15, 1998. In addition, the Division shared its analyses of the operation of the circuit breaker on October 27 with the staffs of the other agencies comprising the President's Working Group on Financial Markets ("Working Group"). The Division's findings regarding circuit breakers are discussed below and were reflected in the Working Group Staff Report on Circuit Breakers, issued on August 18, 1998.

Overall, the Division's survey of the securities markets' systems performance on October 27 and 28 found that the operational capacity enhancements undertaken over the last ten years enabled the markets to accommodate the sharp increases in price volatility and trading volumes on these days with minimal delays and disruptions. While some small problems developed in stock and option quotation systems and Nasdaq, the Division found no indications of the types of large scale breakdowns in automated trading systems that had overwhelmed the markets in October 1987. In terms of broker-dealer systems, the staff found that the firms need to do more to evaluate, expand, and test their systems capacities for peak trading periods. In addition, the staff found that online brokerage firms need to do more to improve their customers' access through the Internet, to respond promptly to customer complaints, and to educate customers about the types of access problems that can develop during peak periods and access alternatives that are available under these circumstances.

Table of Contents

Part II - Findings From Trading Reconstruction

Part III - Determinations Regarding Circuit Breakers

Part IV - Determinations Regarding Operational Capacity

Part V - The Division's Trading Reconstruction

Part I: Introduction and Overview

October 27 was also the first time that the cross-market trading halt circuit breaker procedures had been used since their adoption in 1988. At 2:36 p.m. on October 27,2 the DJIA had declined a total of 350 points from the previous trading session's closing value. This decline triggered a 30-minute halt on the stock, options, and index futures markets. After stock trading resumed at 3:06 p.m., prices fell rapidly to reach the 550-point circuit breaker level at 3:30 p.m., thereby ending trading 30 minutes prior to the normal stock market close.3

On Tuesday, October 28, market prices initially resumed their sharp decline. By 10:06 a.m. the DJIA had declined a total of 187.86 points (2.62%). The market subsequently rallied sharply, with the DJIA closing up 337.17 points (4.71%) at 7498.324 on then-record share volumes of over a billion shares each on the NYSE and the Nasdaq Stock Market ("Nasdaq").5

Throughout these two days, the Commission staff monitored both U.S. and foreign market developments on a 24-hour basis. The staff consulted repeatedly with officials at the U.S. securities markets and clearing organizations to ensure that no systemic problems were arising from the sharp price swings and trading volume surges. Specialized teams of Commission staff also checked to determine if problems were developing in automated trading and price reporting systems, clearance and settlement operations, and market-making in the securities markets, as well as in the financial condition of key broker-dealers. We were also in contact with officials at the other agencies in the President's Working Group on Financial Markets: the Department of the Treasury, the Board of Governors of the Federal Reserve System ("Federal Reserve"), and the Commodity Futures Trading Commission ("CFTC").6 In addition, senior Commission staff consulted with international securities regulators to discuss conditions in other markets. Based on these preliminary reviews, the Commission staff determined that our markets had operated relatively well on October 27 and 28, and that, although some difficulties with automated systems were noted, systemic stress was not evident.

Immediately following these preliminary assessments, Chairman Levitt requested that the Division of Market Regulation ("Division") undertake a more thorough analysis of the markets on October 27 and 28. In particular, the Division was requested to conduct a reconstruction of trading on these days and to determine the effects of the circuit breaker procedures on market price movements. This review sought to determine how the circuit breaker procedures then in effect influenced trading decisions by key market participants on October 27. The Commission's Office of Economic Analysis ("OEA") helped analyze the effects of circuit breakers on the velocity of market price movements and the quality of market-making on October 27.7

In order to perform a timely reconstruction of trading on October 27 and 28, the Division reviewed trading and clearing records for the 30 stocks in the DJIA and the ten largest Nasdaq stocks in terms of capitalization (referred to collectively as the "sample stocks"). We then used a system referred to as the "electronic blue sheet" system8 to obtain information concerning how much customer and proprietary accounts purchased or sold the sample stocks on October 27 and 28 at the most active clearing firms. This trading information was analyzed to identify the largest buyers and sellers in the sample stocks. The information was then cross-checked against information supplied by the Depository Trust Company ("DTC") concerning its most active institutions in terms of net purchases and sales of the sample stocks. These analyses were supplemented by two additional sources of trading information. First, the detailed program trading reports submitted by NYSE member firms were analyzed for October 27 and 28.9 Second, reports provided by the CFTC staff concerning large traders in active stock index futures who were the largest net buyers and sellers in these contracts on October 27 and 28 were reviewed. Based on these trading reviews, the Division interviewed traders at the twelve most active broker-dealers and the three most active mutual fund complexes to discuss in detail how their order flow on October 27 and 28 was influenced by circuit breakers, systems problems, constraints in market liquidity, and other factors.

In addition to the trading reconstruction, the Division undertook an extensive survey of how well the securities markets' automated systems for trading, price-reporting, and clearance and settlement systems operated during the sharp market swings and heavy trading volumes experienced on October 27 and 28. The review analyzed the types of operational problems that developed during this period and whether increases in capacity or other systems enhancements are needed to respond to future market contingencies.

In conjunction with the Division's review, the Commission's Office of Compliance Inspections and Examinations ("OCIE") conducted a series of inspections of full-service and online broker-dealers' trading system capacity for October 27 and 28. These inspections included reviews of customer complaints, which included problems accessing online systems on these trade dates.

While the specific findings and recommendations from the Division's review and OCIE's inspections are non-public, a general overview of these reviews is provided in Part IV this report.

II - Findings From Trading Reconstruction

I. Overview

II. Specific Findings Regarding Types of Traders and Trading Strategies

A. Largest Traders on October 27

B. Trading by Mutual Funds

C. Overall Program Trading

D. Index Arbitrage

The relative reduction in index arbitrage trading on October 27, 1997, is consistent with longer term trends reflected in weekly program trading reports issued by the NYSE over the last few years. These reports indicate that the percentages of index arbitrage program trading have generally declined even as overall program trading has been increasing as a percentage of total NYSE volume.14 While the reduction in index arbitrage may be related to the strictures imposed by NYSE Rule 80A(c),15 this may also be a function of other factors, such as the recent increase in the non-arbitrage program trading strategies discussed below.

E. Quantitative Program Trading Strategies

The firms using quantitative program trading strategies indicated during our interviews that, while they were net sellers of stocks on the afternoon of October 27, they adhered to their models relatively closely and neither the magnitude of the market decline nor the approach of the circuit breaker thresholds caused them to increase the size or frequency of their sell programs. In other words, the likelihood that trading would be halted due to circuit breakers had not been factored into their quantitative models and could not be quickly formulated and inserted into their trading strategies on the afternoon of October 27.

F. "Risk" or "Blind Bid" Programs

The firms employing this strategy maintain that, because the precise makeups of the baskets to be traded are not disclosed by the funds, the firms do not attempt to buy or sell specific securities in anticipation of the later trades. Instead, the firms may buy or sell stock baskets that generally conform to the profiles identified by the funds in order to position the firms for the later trades. Although the firms using this strategy on the afternoon of October 27 were large net sellers of the sample stocks, the firms maintain that this strategy is designed to be "market neutral" because their sales in one portfolio are usually offset by purchases in another portfolio. The firms using these program trading strategies also indicated that, while they were net sellers of the sample stocks on the afternoon of October 27, neither the magnitude of the market decline nor the approach of the circuit breaker thresholds caused them to increase the size or frequency of their sell programs.

G. Dynamic Hedging

H. Index Futures

Part III - Determinations Regarding Circuit Breakers

I. Background

A. The October 1987 Market Break and the Adoption of Circuit Breakers in 1988

The combination of historic price swings and unprecedented trading volumes during October 1987 overwhelmed the operational capacities and liquidity of the securities and futures markets. During October 19, there were frequent delays in reporting quotes and transactions. This contributed to the stress attendant to a 23% price decline. By mid-day on October 20, heavy selling pressure had produced large order imbalances and numerous ad hoc trading halts in individual stocks. Liquidity and pricing difficulties also resulted in uncoordinated mid-day trading suspensions on major options exchanges and several large stock index futures exchanges. In addition, amid rumors that some clearinghouses and several major market participants had failed, a widespread credit breakdown appeared possible. While the subsequent rally in market prices in the afternoon averted more widespread financial problems, the near shutdown of the markets on October 20 became a central focus of several studies of the October 1987 Market Break and resulted in the adoption of circuit breaker procedures in 1988.

Immediately following the October 1987 Market Break, the Presidential Task Force on Market Mechanisms was established with Nicholas F. Brady as Chairman. The report issued by the Task Force on January 8, 1988 ("Brady Report"), recommended a number of initiatives to address future periods of extreme market volatility, including the implementation of circuit breaker mechanisms coordinated across the markets for stocks, options, and stock index futures. The Brady Report noted that the market disorders of October 1987 "became, in effect, ad hoc circuit breakers, reflecting the natural limits to market liquidity." Accordingly, the Brady Report maintained that the October 1987 Market Break "demonstrates that it is far better to design and implement coherent, coordinated circuit breaker mechanisms in advance, than to be left at the mercy of the unavoidable circuit breakers of chaos and system failure."19

After the issuance of the Brady Report, the President's Working Group on Financial Markets ("Working Group") was formed with the mandate to determine the extent to which coordinated regulatory action was necessary to strengthen the nation's financial markets.20 The May 1988 Interim Report ("Interim Report") of the Working Group recommended a number of initiatives to assist the markets in coping with future periods of extraordinary price swings and volume surges. These initiatives included expansion of the operational capacity of the markets, streamlining of clearance and settlement operations, and the adoption of circuit breakers that would provide coordinated trading halts and reopenings for large, rapid market declines that threaten to create panic conditions.

According to the Interim Report, circuit breakers were designed to substitute planned trading halts for ad hoc, destabilizing market closings. Such ad hoc trading halts were manifest during the October 1987 Market Break through systems breakdowns, reduced liquidity, and concerns over trading because of fears of counter-party and clearing corporation failure.21 Thus, the Working Group recommended that all U.S. markets for stocks, options, and futures halt trading for one hour if the DJIA declined 250 points from its previous day's closing level and halt trading for two hours if the DJIA declined 400 points from its previous day's closing level.22 In addition, the Working Group anticipated quarterly reviews of the circuit breaker thresholds to determine whether changes in index levels necessitated changes to the triggers so that they continue to reflect percentage declines approximately equivalent to 12% and 20%.23

Partly in response to the October 1987 Market Break and the recommendations of the Brady Report and the Working Group, the securities and stock index futures markets submitted proposals to the Commission and CFTC in 1988 to implement circuit breakers that would impose temporary trading halts following significant market declines. The circuit breaker rules for the securities and stock index futures markets were implemented on a pilot basis beginning in October 1988,24 and have been extended on that basis each year since their adoption.

The circuit breakers approved in 1988 provided for a one-hour trading halt in all securities markets if the DJIA declined 250 points from its previous day's closing level and for a subsequent two-hour trading halt if the DJIA declined 400 points from its previous day's close. In addition, the original circuit breaker procedures allowed the markets to use abbreviated reopening procedures to permit trading to reopen before the scheduled closing. The abbreviated reopening procedures were also to be used to establish closing prices if the DJIA reached the 250-point trigger during the last hour, but before the last half-hour, of trading, or if the DJIA reached the 400-point trigger during the last two hours, but before the last hour, of trading.25

In approving the original circuit breakers, the Commission and CFTC noted that the circuit breakers were not an attempt to prevent markets from reaching new price levels, but an effort by the securities and futures markets to arrive at a coordinated means to address potentially destabilizing market volatility along the lines of the historic decline of the October 1987 Market Break.26 While concurring in the rationale of the Brady Report and the Interim Report regarding the purpose of circuit breakers, the Commission also believed that circuit breakers would help promote stability in the equity and equity-related markets by providing for increased information flows and enhanced opportunity to assess information during times of extreme market movements. The Commission believed that circuit breakers could provide market participants with an opportunity to re-establish an equilibrium between buying and selling interest and ensure that market participants had a reasonable opportunity to become aware of and respond to a dramatic market decline.27

B. Modifications to the Circuit Breakers in 1996 and 1997

When these circuit breaker changes were approved in July 1996, the agencies reiterated their strong concern that the sharp rise in stock prices since 1988 had reduced the circuit breaker thresholds to levels that were too low. When the circuit breakers were adopted in 1988, the 250-point threshold represented a DJIA decline of 12% and the 400-point threshold represented a decline of 19%. By July 1996, the 250-point and 400-point triggers represented DJIA declines of 4.5% and 7%, respectively. Accordingly, in approving the shorter trading halts in July 1996, the Commission encouraged the markets to continue to evaluate the trigger levels for the trading halts to increase them to levels that reflected their original design.30

Subsequently, when the agencies approved a six-month extension of the circuit breakers in October 1996,31 the Commission again urged the markets to reach a consensus on the size of increases in the trigger levels required to ensure that cross-market trading halts would be imposed only during market declines of historic proportions.32

In response to the agencies' recommendations, the markets submitted proposals to increase the circuit breaker triggers to levels of 350 and 550 points in the DJIA.33 In approving the 350/550 trigger levels through January 31, 1998, the Commission stated that the new trigger levels represented a substantial improvement over the existing 250/400 trigger levels. Nevertheless, the Commission noted that trigger levels should be amended to reflect an extraordinary decline under prevailing market conditions and that the Commission and CFTC would work with the markets to develop procedures for reevaluating the circuit breaker triggers on at least an annual basis.34

II. Findings Regarding the Effects of Circuit Breakers on October 27, 1997

The following discussion is based on the Division's trading reconstruction and OEA's data for October 27, 1997, as well as on information compiled in interviews with the most active broker-dealers and mutual fund complexes.

A. The First Circuit Breaker Was Too Low

Traders indicated that, as the first circuit breaker trigger was being approached, overall trading and market operations appeared to be orderly. While a few firms indicated that pricing in stocks, options, and futures was becoming difficult because of price gapping36 and price data transmission delays, most firms indicated that they were experiencing no significant problems along these lines prior to the 2:36 p.m. halt.37 While some traders reported a reduction in market liquidity, there were no indications that liquidity problems might result in the types of ad hoc trading halts that circuit breakers were designed to address.

No firms reported panic selling by their institutional, retail or professional accounts.38 Several firms indicated that their customer and proprietary accounts, particularly accounts that had been purchasers of stocks earlier in the day, pulled back from trading as the 350-point circuit breaker approached due to the uncertainties surrounding the first-ever triggering of the cross-market halt procedures. Only a few firms indicated that institutions increased somewhat the size and frequency of their outstanding sell orders in order to beat the pending halt.

B. Evidence Is Mixed Whether the First Circuit Breaker Had a "Magnet Effect"

First, the decline in the DJIA came within seven points of the 350-point trigger at 1:59 p.m., but prices then stabilized and recovered approximately 70 points by 2:10 p.m. Second, there is no clear pattern of an accelerating market decline from 2:10 p.m. to 2:36 p.m. as the DJIA reached the trigger point. As indicated in the graph at B-5, the Division found that the largest one-minute percentage decline in the DJIA during this period occurred around 2:16 p.m. The price decline abated for a few minutes shortly after 2:20 p.m. and again at 2:34 p.m. before the 350-point threshold was reached.

Second, although the OEA data indicate that the rate of decline in S&P 500 stocks accelerated as the first circuit breaker approached, with S&P 500 stocks declining at a rate of 0.03% per minute between 1:03 p.m. and 2:35 p.m. (compared 0.01% per minute between 9:30 a.m. and 1:03 p.m.), the rate of decline is not as dramatic as the rate of decline between 3:06 p.m. and 3:30 p.m., when S&P 500 stocks declined at a rate of 0.10% per minute.39

Third, there was no appreciable increase in trading volume on the NYSE in the period immediately prior to the first circuit breaker.40 If there had been a magnet effect, a surge in trading volume would have been expected as a function of investors rushing into the market in an attempt to trade before the circuit breaker could be implemented.

Overall, therefore, the Division believes that the evidence is mixed as to whether the 350-point circuit breaker had a magnet effect.

C. The 30-Minute Halt Was Benign but of Little Utility

In addition, most firms did not find the halt to be useful in assessing market conditions or in handling inquiries from institutional customers. In fact, the NYSE disseminated pre-opening indications in only a few stocks because most stocks had no sizeable order imbalances prior to the 3:06 p.m. reopening. Overall, it appears that the vast majority of firms simply had their trading staffs wait out the halt. Traders at a few firms were of the opinion that the halt may have somewhat exacerbated worries among professional and institutional traders. Most firms, however, found the halt to have been benign.

D. Market Reopenings Appear To Have Been Orderly

Firms generally experienced no significant difficulties in executing orders in listed or Nasdaq stocks immediately following the halt. Several firms, however, commented on the overall reduction in liquidity in the stock, options, and futures markets from 3:06 p.m. to 3:30 p.m. as buyers withdrew from the markets. In general, the rapid reopening of NYSE and Nasdaq stocks after the first circuit breaker was probably due to the fact that the first halt occurred after only a moderate decline while trading was still orderly.

E. The Second Circuit Breaker May Have Had a Magnet Effect

In addition, between 3:06 p.m. and 3:30 p.m., the price decline did not accelerate in a clear pattern. As indicated in the graph at B-6, the largest one-minute percentage declines in the DJIA during this period were experienced around 3:12 p.m. and 3:14 p.m. and again around 3:24 p.m. and 3:25 p.m. and the rate of decline abated somewhat in the intervening period and again just before the 550-point threshold was reached at 3:30 p.m. The absence of a clear pattern of an accelerating market decline from 3:06 p.m. to 3:30 p.m. again is not entirely consistent with a magnet effect for the second circuit breaker level.

Nevertheless, the velocity of the market decline and the OEA market-making statistics for the period immediately before the second circuit breaker halt indicate that it may have had some magnet effect.

During the period between the reopening of the markets at 3:06 p.m. and the triggering of the second circuit breaker at 3:30 p.m., the DJIA declined over 200 points. The OEA data indicate that the velocity of the price decline in S&P 500 stocks also increased significantly during this period, with S&P 500 stocks declining at a rate of 0.10% per minute (or 6% per hour) between 3:06 p.m. and 3:30 p.m., ten times more quickly than their decline at a rate of 0.01% per minute (or 0.6% per hour) between 9:30 a.m. and 1:03 p.m. The price decline for the period immediately before the second circuit breaker also is more rapid than the decline between 1:03 p.m. and 2:35 p.m., when S&P 500 stocks declined at a rate of 0.03% per minute (or 1.8% per hour).42

The increase in quote spreads between 3:06 p.m. and 3:30 p.m. also suggests a possible magnet effect associated with the second circuit breaker. Between 3:06 p.m. and 3:30 p.m., mean relative spreads (the quoted dollar bid-ask spread divided by the spread mid-point) for S&P 500 stocks were approximately 46 basis points, a 47% increase over the mean relative spread of 31 basis points on October 23.43

Effective quote spreads (calculated by doubling the difference between the trade price and the midpoint of the bid-ask spread), which reflect the cost of trades executed inside the quote spread, also increased throughout the day on October 27, but most significantly after the first circuit breaker. For S&P 500 stocks traded on the NYSE, the mean effective spread was 10.6 cents per share between 9:30 a.m. and 1:03 p.m., 12.9 cents per share between 1:03 p.m. and 2:35 p.m., and 18.1 cents per share between 3:06 p.m. and 3:30 p.m. Mean effective quote spreads for DJIA stocks also increased during these periods, rising from 10.4 cents per share during the first period, to 14 cents per share during the second period, to 23.4 cents per share during the third period.44

Finally, the ratio of bid depth to ask depth exhibited a similar pattern. The median bid depth to ask depth ratio fell somewhat from the morning session to the first pre-halt period on October 27, then dropped substantially in the post-halt period.45

Although the sharp increase in the rate of decline in the S&P 500 stocks between 3:06 p.m. and 3:30 p.m. is consistent with a magnet effect for the second circuit breaker, it is not possible to state definitively, on the basis of a single event, that the second circuit breaker produced a magnet effect. Given the increase in volatility prevailing at that time, it is impossible to place responsibility for the swiftness of the decline between 3:06 p.m. and 3:30 p.m. solely on the second circuit breaker.

F. A Normal Close Should Have Been Permitted

The OEA data for October 27 are consistent with the traders' observation that the early market closure was unnecessary. Although quote spreads widened throughout the day on October 27, other measures of market quality suggest that, overall, the markets were functioning in an orderly manner with sufficient liquidity on October 27.

For example, the number of transactions and shares traded at the bid before a downtick (i.e., a change in price downward) reflect the amount of liquidity at a bid quote and the markets' ability to absorb selling pressure. On October 27, data in both S&P 500 stocks and DJIA stocks reveal a fair amount of liquidity at each quote and suggest that the markets were able to function in an orderly manner. The OEA data in Attachment D do not indicate that there was chaotic pricing or destabilizing price moves in the late afternoon of October 27. Although certain measures, such as the ratio of bid depth to ask depth, quote spreads, and acceleration of price declines, show a deterioration after the first circuit breaker, they clearly do not indicate an impending systemic breakdown or failure to maintain an orderly market. Indeed, the decline in market measures may have been due not only to the increase in volatility, but also in part to the uncertainty caused by the prospect of a premature close of trading from the approach of the second circuit breaker.

G. Most Firms Did Not Have Large Unexecuted Orders

While the number of unexecuted orders when the markets closed at 3:30 p.m. apparently was not large at the firms interviewed by the Division, OCIE's inspections of broker-dealers found that there was confusion at firms as to how to properly execute pending customer orders and orders that were received after the markets closed on October 27. Most firms apparently executed these orders when the stock markets reopened on October 28. In order to address uncertainties if a future market decline results in a circuit breaker halt that closes the markets early, the Division has discussed its views in a staff legal bulletin which has been issued concurrently with the release of this report.47 The bulletin reminds broker-dealers that customer orders received under these circumstances should be treated in the same manner as they would be when an SRO imposes a regulatory trading halt in a single security. Such security-specific regulatory halts occur virtually on a daily basis. If an issuer is about to release material corporate news during a trading session, the SRO that serves as the primary market for the issuer's securities would impose a regulatory trading halt (for "news pending") that is honored by all of the other SROs until the news is disseminated and the halt is lifted. Broker-dealers and the markets already know how to handle orders when an individual stock is halted under these circumstances, and the Division believes that there is no reason to adopt different practices for circuit breaker halts. Accordingly, the Division believes that, unless a customer indicates otherwise when submitting an order, orders that are pending when a circuit breaker is triggered, or that are received afterward, should be treated as "Good 'Till Cancelled" orders and held for execution at the reopening on the next trading session. Pending orders marked as "At-the-Close" orders (including "Market-at-Close" or "Limit-at-Close") should be treated as cancelled. In addition, no new orders relating to closing prices may be accepted.48 The bulletin also indicates that, when circuit breakers have closed the markets, broker-dealers should make every reasonable effort to notify customers how their orders will be treated.49

H. The Early Close Could Have Resulted in Severe Derivative-Related Losses

I. Evidence Is Mixed on Whether the Early Close Harmed Foreign Markets

Nevertheless, the Division recognizes that foreign markets have often fallen sharply immediately following previous U.S. market declines, such as those in October 1987,53 October 1989,54 and November 1991,55 when circuit breakers were not employed. It may not be possible, therefore, to determine the precise extent to which the continuing market declines on foreign markets on October 28, 1997, were due to the early market closure on October 27 rather than the severity of the U.S. market decline itself or other factors related to the foreign markets' own trading dynamics. Nevertheless, the severity of the foreign market declines on the morning of October 28 appears to indicate that the early closure of the U.S. market on October 27 may have at least contributed to these declines

J. Preopening Pricing on October 28 Was Viewed As Helpful

III. Regulatory Initiatives Since October 27, 1997

As an interim measure, the markets adopted modest changes designed to reduce the likelihood that the 350/550-point trigger levels would preclude normal market closes. Specifically, the Commission and the CFTC approved changes effective through April 1998 which provided that the markets would not implement the 30-minute circuit breaker halt if the DJIA reached the 350-point trigger on or after 3:00 p.m. and would halt trading for only 30 minutes (rather than an hour) if the DJIA reached the 550-point threshold on or after 2:00 p.m. but before 3:00 p.m. If the DJIA reached the 550-point threshold on or after 3:00 p.m., the markets would have continued to use the one-hour halt, which would have ended the trading session early.56

In discussions with the securities and stock index futures markets aimed at achieving a consensus on expanded circuit breaker levels, the Commission indicated its firm belief that the 10% and 20% circuit breakers should not close the markets prematurely during the trading day. In addition, at a U.S. Senate hearing on January 29, 1998, the Working Group agencies and most senators indicated a strong preference for the markets to remain open whenever possible and a disinclination for circuit breakers to close the markets for the day.

In response to Congress's and the agencies' concerns, the securities and futures markets submitted proposals to revise their circuit breaker procedures. The Commission and the CFTC approved the revised procedures in April 1998.57 The circuit breakers adopted by the securities exchanges establish trading halts following one-day declines in the DJIA of 10%, 20%, and 30%. The NYSE will calculate the trigger levels at the beginning of each calendar quarter, using the average closing value of the DJIA for the previous month to establish specific point values for the quarter. Under the securities exchanges' revised circuit breaker procedures, trading will halt for one hour if the DJIA declines 10% prior to 2:00 p.m., and for one-half hour if the DJIA declines 10% between 2:00 p.m. and 2:30 p.m. If the DJIA declines by 10% at or after 2:30 p.m., trading will not halt at the 10% level. If the DJIA declines 20% prior to 1:00 p.m., trading will halt for two hours; trading will halt for one hour if the DJIA declines 20% between 1:00 p.m. and 2:00 p.m., and trading will halt for the remainder of the day if a 20% decline occurs at or after 2:00 p.m. If the DJIA declines 30% at any time, trading will halt for the remainder of the day.

The futures exchanges trading stock index futures have adopted substantially identical circuit breaker procedures. However, the Chicago Mercantile Exchange ("CME") has adopted a daily price limit for the S&P 500 futures that will permit a maximum daily downward price movement of 20%, while the securities exchanges' circuit breaker procedures will permit trading in the range of 20% to 30% down prior to 2:00 p.m. In addition, the CME's variation margin settlement values will be based on the 20% limit price, rather than on a price derived from the closing index value. While noting the disparities in the markets' procedures and urging the CME to reconsider its 20% cap on variation margin, the regulators, in approving the revised procedures, concluded that the markets' rules are substantially identical for purposes of the effectiveness of the circuit breaker rules.58

In adopting the securities markets' revised circuit breaker procedures, the Commission noted that the amended trigger levels reflect the type of severe one-day market declines that circuit breakers were intended to address. The Commission concluded that the revised trigger levels are consistent with the intended design and function of the circuit breakers, and that they should not cause premature or unnecessary trading halts. In addition, the Commission found that the revised circuit breaker procedures sufficiently address the need for the markets to remain open or to reopen during the trading day to permit an orderly market close.59

IV. Conclusions Regarding Cross-Market Trading Halts

First, the circuit breaker thresholds needed to be raised significantly from those in place on October 27. When the 350-point trigger was reached on October 27, the DJIA was down only 4.54%, a level that had been reached on 11 previous days since 1945. Moreover, there was little evidence of the types of market liquidity constraints that would have justified cross-market halts. Circuit breaker halts should be reserved for an abrupt market decline of a magnitude that raises concerns that the exhaustion of market liquidity might result in uncoordinated, ad hoc market closures.

Second, the trading dynamics on October 27 illustrated the need for circuit breakers to permit trading to resume whenever feasible for orderly market closings. The early market closure on October 27 does not appear to have been necessary. Moreover, investor concerns that the second circuit breaker would close the market may have accelerated the price declines in the last 24 minutes. Concerns over a likely premature market close may have resulted in a transfer of selling pressure to the futures markets (which offered faster executions in the limited time remaining), with stock prices quickly following futures down to the 550-point trigger level. In addition, although the harm to investors and markets from the early closure on October 27 may be difficult to quantify, our review highlighted the types of losses that could result if circuit breakers prevent day-end unwindings of highly leveraged derivative positions. While the derivative losses on October 27 appear to have been isolated and manageable, losses could have been many times worse if unwindings had been precluded on dates when more exchange-traded and OTC derivatives expire.60

Third, the events of October 27 reinforced the need for regulators to periodically re-examine circuit breaker procedures to ensure that they reflect both changing market levels and the increasing capacity of the markets to handle greater trading volumes and price volatility in an orderly manner. As markets continue to grow and change, the regulatory agencies and the self-regulatory organizations must monitor and revise circuit breakers and other protective measures to ensure that they continue to function as intended and achieve their goals with minimal market disruptions.

IV - Determinations Regarding Operational Capacity

I. Market Systems

A. Trading and Clearance and Settlement Systems

Nevertheless, the Division noted that a number of alternative trading systems, known as electronic communication networks ("ECNs"), had their systems capacity severely strained during the sharp price swings and heavy trading volumes on October 27 and 28. Several ECNs experienced system outages and one significant ECN withdrew its quotes from Nasdaq because of lack of capacity. Similarly, a major interdealer broker in non-exempt securities experienced serious capacity problems in processing the large number of transactions on these days and had to close down temporarily.

B. Exchange Quotations for Stocks and Options

C. Nasdaq Systems

II. Broker-Dealer Systems

Overall, the broker-dealers did not have adequate excess capacity to handle peak periods on October 28. As a result, full-service broker-dealers experienced at least some trading system problems on October 28, which interfered with timely executions of customer orders. In addition, while several online broker-dealers executed record numbers of customer orders on October 28, some firms experienced limited degradation in their Internet access systems, which prevented certain customers from accessing their online accounts in a timely manner. While several firms offered price adjustments to customers who received delayed executions, more efforts can be made by broker-dealers to address these problems.

The Division has sought to address the need for improvements in these areas as part of its staff legal bulletin that has been issued concurrently with this report.65 In particular, the firms should make greater efforts to meet the guidelines set forth in the Commission's automation review policy statements. Broker-dealers should recognize the importance of having adequate capacity to handle high volume or high volatility trading days, and to conduct capacity planning on a regular basis.66 Broker-dealers should also consider other appropriate measures to minimize potential capacity problems, including developing back-up technology systems and developing procedures for handing system problems. Online broker-dealers might consider additional steps, such as enhancing their internal capacity and their links to the Internet, prioritizing customer access at peak times (such as limiting some general price inquiries in order to facilitate executions of customer orders), educating their customers about Internet access issues, and providing customers with alternative means to place orders when Internet access is slow or unavailable.67 Broker-dealers should also use every reasonable effort to notify customers about operational difficulties.68

Part V - The Division's Trading Reconstruction

I. Market Context

A. The Bull Market

Much of the gain in share prices was accompanied by generally declining interest rates, particularly since late 1994. For example, while yields for the benchmark 30-year Treasury bonds fluctuated throughout this period, bond yields generally declined from a high of 8.166% on November 7, 1994, to around 6.3% on July 31, 1997, with the yield edging up to 6.481% as the DJIA peaked on August 6.73 This environment of relatively low interest rates, coupled with indications of continuing low inflation and rising corporate earnings, was cited by many market commentators as justifying stock price valuations that were high by some traditional valuation measures.74 For example, by September 1997, the dividend yield on the S&P 500 stocks had reached a then-historic low of 1.63%, well below the yield of 2.83% in July 1987 as stock prices were peaking before the October 1987 Market Break.75 Similarly, the price/earnings ratio for S&P 500 stocks reached a peak of 24.00 in September 1997, exceeding the 22.33 ratio in August 1987.76 Nevertheless, even as share prices pulled back slightly for several weeks following the August 6 high, many investors and market commentators continued to view the favorable interest rate, inflation, and corporate earnings environment as justifying further share price gains.

The continued optimism for stock prices was also apparently fueled by the historic increase in investor interest in equity mutual funds since the late 1980s. According to the Investment Company Institute, the asset values of equity funds had soared from $76 billion in January 1984 to over $2.39 trillion in September 1997, an increase of 3,030%.77 Strong investor interest in equity funds evidently continued even after short-term stock price declines, such as the 247.37 point (3.11%) decline in the DJIA on August 15, 1997.78 The high level of interest in U.S. stocks was also reflected in the increasing share volume levels on the NYSE and Nasdaq. The NYSE average daily share volume grew from less than 200 million shares in 1991 to over 500 million shares by October 1997. During the same period, Nasdaq average daily share volumes rose to over 600 million shares.79 Volume levels on the NYSE and Nasdaq remained strong in 1997 through October.80

Nevertheless, a number of market commentators expressed increased concern in the late Summer and early Fall of 1997 that high stock valuations made the market vulnerable to a sharp pull-back in prices if one or more factors arose, such as a changed interest rate outlook by the Federal Reserve, signs of incipient inflation, disappointing corporate earnings or unexpected world events such as a war or other international crisis.81 These concerns were particularly strong as the ten-year anniversary of the October 1987 Market Break approached.82

B. The Decline in Pacific Rim Markets

This pattern remained in place through much of 1997, when U.S. share prices continued to rise despite mounting losses on some of the smaller Southeast Asian markets. For example, the stock market in Thailand, as measured by the Bangkok SET Index, had a year-to-date decline of more than 44% by June 19, 1997. U.S. stock prices continued to rise during this period.85 This relationship was maintained even through July and August 1997 as the currency crisis in Thailand accelerated.86 Similarly, U.S. stock prices showed little reaction as currency volatility contributed to declines in other small Southeast Asian markets, such as Indonesia and Malaysia, which experienced year-to-date percentage losses in their stock markets that reached almost 23% and 35%, respectively, by late August 1997.87 In fact, prices for broad U.S. indexes, such as the S&P 500 index, established new closing highs on October 7, 1997, despite the sharp declines in the smaller Asian stock markets over the preceding months.

Anxieties over possible "contagion" effects began to be reflected in stock markets worldwide, however, when the currency and share price instability appeared to spread to other Pacific Rim markets, such as the much larger Hong Kong market, in the week of October 20, 1997. On the three trading sessions on Monday, October 20, to Wednesday, October 22, Hong Kong share prices declined 14.43% in reaction to concerns over the ability of the Hong Kong Dollar to remain "pegged" to the U.S. Dollar. Despite the severity of this three-day decline in Hong Kong, however, the reaction in other markets began to build only slowly.88

This situation began to change when short-term interest rates in Hong Kong were raised sharply on Thursday, October 23, in order to support the Hong Kong Dollar, and the decline in share prices in Hong Kong accelerated. When the Hong Kong Hang Seng Index plunged 1,211.47 points (10.41%) to close at 10,426.30 on October 23, the Tokyo Nikkei closed down 536.06 points (3.03%) at 17,151.55, the London FTSE closed down 157.30 points (3.06%) at 4,991.50, and the DJIA declined 186.88 points (2.33%) to close at 7,847.77. Comparable price declines were experienced on other markets worldwide.89

On Friday, October 24, the Hang Seng Index rallied 718.04 points (6.89%) to close at 11,144,34, and the Nikkei closed up 212.19 points (1.24%) at 17,363.74, while the FTSE declined only 21.30 points (0.43%) to close at 4970.20.90 U.S. stock prices initially rallied on the morning of October 24, with the DJIA rising almost 92 points at its intra-day high. Continuing worries over the effects of the Asian turmoil on the U.S. bull market, however, resulted in a further decline in U.S. share prices throughout the afternoon. As a result, the DJIA closed down 132.36 points (1.69%) at 7,715.41. By the end of the trading session on October 24, it was also becoming clear to investors that the market turmoil was taking a severe toll on other emerging markets, such as stock markets in Latin America, several of which had been outperforming U.S. markets for much of the year.91 For example, in Brazil, the Sao Paulo Bovespa Index closed down 2.97% on October 24, while the Argentine Merval Index fell 4.17% and the Mexican I.P.C. All-Share Index fell 2.74%.

The expansion of the market turmoil beyond small Southeast Asian markets to larger Pacific Rim markets and emerging markets in Latin America during the week of October 20 evidently raised serious concerns among institutional and professional traders that economic slowdowns and even deflationary pressures in these markets could adversely affect U.S. corporate earnings. Moreover, if earnings projections were adjusted downward significantly, these investors evidently had reservations that the historically high U.S. stock valuations at that time were likely to be surpassed or even sustained in the short term.92 As a result, by Monday, October 27, some institutional and professional traders were prepared to reduce their equity exposure in the U.S. or at least to increase their equity hedges significantly, either directly through stock sales or indirectly through trades in stock index futures.

II. October 27, 1997

A. Foreign Market Declines and Pre-Opening Indications in the U.S.

On Monday, October 27, the Hong Kong Hang Seng Index declined 646.14 points (5.80%) to close at 10,498.20, thereby reversing much of Friday's gain. While the reaction to this renewed decline in Hong Kong was moderate in Tokyo, where the Nikkei closed down only 325.38 points (1.87%) at 17,038.36, the overnight share price declines in other markets were more severe.93 By around 9:00 a.m. on October 27, the decline in European stock markets was already widespread. While the London FTSE was trading down a moderate 1.53%, the Frankfurt Dax had closed down 4.24% (after-hours trading on the Frankfurt IBIS Dax was showing that index down 2.71%), and the Paris CAC was trading down 3.09%.94

Nevertheless, the pre-opening indications in the U.S. markets remained fairly moderate on the morning of October 27. The December S&P 500 futures contract on the Chicago Mercantile Exchange's Globex automated trading system had traded as low as down 14.00 points (1.48%) at 930.00 around 7:55 a.m. The futures, however, ended the after-hours trading session at 9:00 a.m. down only 10.25 points (1.09%) at 933.75, which forecast a decline in the DJIA of approximately 80-to-100 points at the 9:30 a.m. opening of the NYSE.95

When the S&P 500 futures began their regular floor trading session at 9:15 a.m.,96 prices actually improved somewhat from the levels set earlier on the Globex automated trading system. While the futures had ended the Globex session at 9:00 a.m. down 10.25 points, floor trading began with the futures down only 8.00 points (0.85%) at 936.00. Futures prices fluctuated in approximately a 2 1/2 point range as the 9:30 a.m. opening for the stock market approached. As of 9:30 a.m., the S&P 500 futures were down 7.00 points (0.74%) at 937.00, which forecast a decline in the DJIA of 50-to-70 points.97

B. Market Declines in the Morning

Therefore, even at the low point for the DJIA on the morning of October 27, the decline in the index was not particularly severe in percentage terms. On an intra-day basis, the 2.37% decline in the DJIA at 11:30 a.m. on October 27 had been exceeded on six previous trading sessions in 1997 alone.106 Similarly, the CFTC staff reports that the 15-point price limit in the S&P 500 futures, which was triggered at 11:14 a.m. on October 27, had been reached on 15 prior trading sessions in 1997.

The relatively moderate nature of the market decline on the morning of October 27 is also reflected in the price velocity statistics compiled by OEA. These statistics show that the rate of decline in the S&P 500 index was only 0.01% per minute in the interval from 9:30 a.m. to 1:03 p.m. on October 27, compared to a rate of 0.007% per minute for the same period on October 23.107 This pattern is also reflected in the quotation and market-making statistics developed by OEA. While OEA calculated that the quote percent spreads for NYSE and Nasdaq stocks from 9:30 a.m. to 1:03 p.m. on October 27 were somewhat wider than those during a comparable period on the control date of October 23, 1997, the October 27 morning spreads were significantly narrower than those evident during the late afternoon.108 The NASD staff reported comparable quote spread statistics for Nasdaq stocks. Specifically, the NASD staff calculated that, while Nasdaq spreads were 11% higher than normal on the morning of October 27, they were significantly narrower than in the afternoon, when Nasdaq spreads were running 17% to 19% above normal. The OEA statistics for the "effective spreads" for NYSE stocks on the morning of October 27 also generally followed this pattern.109

In another measure of liquidity, OEA found that the "bid down tick" trades and volumes for NYSE stocks on the morning of October 27 were generally somewhat greater than those during a comparable period on the October 23 control date.110 On the other hand, OEA found that the average bid and ask depths (greater depths, in general, indicate greater liquidity) were somewhat less in key NYSE stocks on the morning of October 27 as opposed to those on the morning of October 23.111

In terms of NYSE specialist activity, OEA found that specialist trading patterns on October 23 and 27 were similar. On both days, specialists in aggregate were slight net sellers of DJIA stocks from 9:30 a.m. to 1:03 p.m. and slight net buyers of stocks in the S&P 500 index.112 In view of the up and down spikes in share prices throughout the morning and early afternoon on October 27, it is not surprising that specialist buying and selling during this period were more or less evenly matched.

It appears that the stock market decline on the morning of October 27 was accompanied by only moderately heavy trading volume. The Division found that 75.4 million shares traded on the NYSE by 10:00 a.m., which was only 6.38% higher than the first-half hour average for the previous week. Even by 10:30 a.m., total NYSE volume was only 140.1 million shares, which was only 7.21% above the first-hour average for the prior week, and well below the first-hour volume then-record of 279 million shares set on December 15, 1995. For the morning as a whole, NYSE share volume totalled only 318.1 million shares, a level only 15.41% above the prior week's average.113

Overall, the relatively moderate price declines and trading volume levels on the morning of October 27 are consistent with the indications of order flows that were provided to the Division in interviews with the most active trading firms and mutual fund complexes. Most of the traders indicated in their interviews that there were no signs of panic selling by institutional or proprietary accounts in the morning despite the severity of the price declines in foreign markets overnight. The frequent periods throughout the morning in which U.S. share prices stabilized temporarily and even staged partial recoveries are also consistent with reports from traders that a number of institutional accounts took advantage of some of the morning's price declines to step in and buy stocks that appeared oversold. This is also confirmed by the indications that the strongest relative increases in NYSE share volume on the morning of October 27 (when compared to trading levels in the prior week) occurred as stock prices staged a partial recovery from 11:30 a.m. to 12:00 p.m.

The Division found that the levels of program trading on the morning of October 27 were also more or less normal. The graphs at B-12 and B-13 indicate that, while there were repeated surges in program selling throughout the trading session on October 27, most of the program buying appears to have been concentrated in the morning and early afternoon.114 As a result, many of the periods in the morning when share prices stabilized and managed partial recoveries coincided with periods when program buying outpaced program selling. For example, while sell orders accounted for 69% of all program orders when share prices fell sharply in the first ten minutes of trading,115 this situation was quickly reversed as share prices recovered. Buy orders accounted for 89% of program orders from 9:40 to 9:50 a.m. and 92% of program orders from 9:50 to 10:00 a.m. This pattern was particularly pronounced when share prices began their strongest recovery in the morning between 11:30 a.m. and 11:40 a.m., when 95% of all program trading orders were on the buy-side.116 The statistics compiled by OEA that provide an account type breakdown for NYSE volume in S&P 500 index stocks also reflect the slight preponderance of buy-side activity by program trading firms during this interval.117

C. Market Declines to the First Circuit Breaker at 2:36 p.m.

Although by 1:59 p.m. the decline in the DJIA had come within seven points of the 350-point trigger for the first circuit breaker halt, prices rebounded approximately 70 points by 2:10 p.m., leaving the DJIA down 273.33 points (3.54%). A similar partial price recovery was experienced in the S&P 500 index, which reduced its decline to 34.69 points (3.68%) at 2:10 p.m. Nasdaq stocks also followed this trend to reduce their decline to 84.82 points (5.14%) by 2:11 p.m.

Share prices thereafter resumed their decline. The DJIA was down 327.84 points (4.25%) at 2:20 p.m. The DJIA recouped over 20 points in the next few minutes, and fell once again to reach minus 354.36 points (4.59%) at 2:36 p.m., thereby triggering the first circuit breaker trading halt. At that point, the S&P 500 index was down 43.63 points (4.63%) and the Nasdaq composite index was down 92.86 points (5.62%).119 The S&P 500 futures had hit their 45-point price limit at 2:35 p.m., which resulted in a trading halt in the futures when the first circuit breaker was implemented.120

Between 2:00 p.m. and 2:30 p.m., share volume continued to be strong, with 60.5 million shares trading on the NYSE, which represented an increase of 74.25% above the prior week's average for this period. Nevertheless, there was no appreciable increase in volume on the NYSE immediately prior to the circuit breaker triggering. Specifically, the NYSE reported that 17.7 million shares traded from 2:30 p.m. to 2:36 p.m., compared to 18.4 million shares traded during the previous ten minutes.

The statistics developed by OEA highlight the increase in the velocity of the stock market decline in the period leading up to the first circuit breaker triggering at 2:36 p.m. Specifically, OEA calculated that the rate of the price decline in the S&P 500 index increased from 0.01% per minute in the interval from 9:30 a.m. to 1:03 p.m. to a rate of 0.03% per minute from 1:03 p.m. to 2:35 p.m.121

Quote spreads widened as the market decline accelerated. The quote spreads for NYSE and Nasdaq stocks from 1:03 p.m. to 2:35 p.m. on October 27 were significantly wider than those during a comparable period on the control date of October 23. Moreover, the quote spreads had widened from those evident earlier on October 27.122 Similar patterns were found for the effective spreads for NYSE stocks from 1:03 p.m. to 2:35 p.m. on October 27.123

Nevertheless, the markets were still functioning in an orderly manner with sufficient liquidity and operational capacity. For example, OEA found that the statistics for "bid down tick" trades and volumes for S&P 500 index and DJIA stocks generally showed slightly more depth in the interval from 1:03 p.m. to 2:35 p.m. on October 27 in comparison with the statistics for the earlier part of the day.124 In terms of NYSE specialist activity, OEA found that specialists in aggregate had increased their net buying in both DJIA and S&P 500 index stocks as the share price declines accelerated.125 The price declines were increasing in velocity, but prices were not gapping or exhibiting poor continuity.

The price trends found by the Division and OEA for the period leading up to the first circuit breaker halt are generally consistent with the indications of order flows that were provided to the Division in interviews with the most active trading firms and mutual fund complexes. In particular, the rebounds in share prices around 2:00 p.m. and again around 2:20 p.m. are consistent with reports from traders that a number of institutional accounts continued to take advantage of the price declines in the early afternoon to step in and buy stocks that appeared oversold. The trading volumes and the market depth statistics for this period, however, are somewhat stronger than the market conditions indicated by the traders in their interviews. As discussed above, while a few traders acknowledged that some customers increased their selling as the market decline worsened, many traders claimed that selling by their institutional and proprietary accounts abated somewhat as the circuit breaker threshold was approached. Moreover, a number of traders observed that, while trading was orderly, there were already some signs of a general deterioration in market liquidity and depth as the price declines approached the first circuit breaker trigger point.

The Division found that program selling reached significant levels during several periods of the sharpest market declines from 1:50 p.m. to 2:36 p.m.126 For example, 98% of the program trading from 1:50 p.m. to 2:00 p.m. was on the sell-side, with the 6.47 million shares of program selling representing 21.53% of total NYSE volume and 34% of NYSE volume in S&P 500 stocks. Similarly, 97% of the program trading from 2:30 p.m. to 2:36 p.m. was on the sell-side, with the 4.74 million shares of program selling representing 26.79% of total NYSE volume and 40.20% of NYSE volume in S&P 500 stocks. Nevertheless, there were also periods during the market declines after 2:00 p.m. in which program trading was negligible. For example, program selling accounted for only 5.47% of total NYSE volume and 9.02% of NYSE volume in S&P 500 stocks from 2:20 p.m. to 2:30 p.m. In addition, while the statistics developed by OEA concerning program trading in the interval from 1:03 p.m. to 2:35 p.m. show an increase in program sell volume in the S&P 500 stocks (in comparison with program selling in the earlier part of the day), the percentage of share volume from program selling remained a moderate 18.6%.127 Overall, therefore, it appears that, while program selling may have contributed to the velocity of the stock market declines at some key periods, this type of selling was insufficient to explain the magnitude of the market declines leading to the first circuit breaker triggering at 2:36 p.m.

Our trading analysis also indicates that most of the program selling that was effected during the market declines leading to the circuit breaker triggering did not involve classic index arbitrage. From 1:50 p.m. to 2:36 p.m., index arbitrage was reported to constitute only 7.93% of total program selling. Another 31.51% of program selling, however, was for quantitative program trading strategies that had some elements of arbitrage timing of futures trading. Nevertheless, 60.56% of program selling during this period was reported to have no arbitrage elements. As discussed earlier, this non-arbitrage program selling was reported to have been effected for a variety of trading strategies, including quantitative and risk strategies, as well as some isolated programs designed to dynamically hedge index positions maintained by firm proprietary accounts.128 In interviews with the most active program trading firms, their traders indicated that their quantitative and risk strategies were not particularly influenced by the severity of the market declines after 1:50 p.m. or the likelihood that the first circuit breaker trading halt would be implemented. Instead, most program traders indicated that they employed much the same types of strategies as they had during the more moderate market declines in the morning. In addition, the traders claimed that they did not have to significantly increase the size or frequency of their programs as the DJIA approached the 350-point circuit breaker level.

D. Market Declines to the Second Circuit Breaker at 3:30 p.m.

When trading resumed in the stock market at 3:06 p.m., it appears that the market reopening was handled relatively smoothly. The vast majority of stocks on the NYSE and Nasdaq reopened quickly with few significant "gaps" between their prices before and after the halts. For example, the DJIA traded in about a seven-point range from 3:06 p.m. to 3:09 p.m. as stocks reopened. As discussed above, virtually all of the firms that were interviewed by the Division viewed the market reopenings after the first circuit breaker halt to have been surprisingly orderly. Moreover, OEA's statistics found that over 82% of the S&P 500 stocks on the NYSE had reopened by 3:10 p.m. and that over 93% of the DJIA stocks had reopened by that time.130

By 3:07 p.m., trading had resumed in stocks representing over 50% of the capitalization of the S&P 500 index, thereby permitting trading to resume in the index futures. At around 3:10 p.m., however, S&P 500 futures fell rapidly in price, falling 15 points in eight minutes. The futures reached their 70-point price limit by 3:25 p.m.131 Stock prices quickly followed the futures lower, with the DJIA reaching the 550-point circuit breaker trigger at 3:30 p.m., thereby ending the regular trading sessions for U.S. stock, options, and index futures markets.132 After all last-sale prices were reported, the DJIA ended the day down 554.26 points (7.18%) at 7,161.15. The S&P 500 index closed down 64.65 points (6.87%) at 876.99. The Nasdaq composite index closed down 115.83 points (7.02%) at 1,535.09.133

Trading volumes remained strong from 3:06 p.m. to 3:30 p.m., with 80.1 million shares trading on the NYSE, representing an increase of 74.81% from the previous week's average for this period.134 Even with the abbreviated trading hours on October 27, daily share volume set a new record on the NYSE and volume on Nasdaq was the then-fourth heaviest on record. Nevertheless, even these trading volumes on October 27 were only moderately larger than recent trading levels. For example, while the October 27 daily volume on the NYSE totalled 685.5 million shares, representing a 33% increase over the year-to-date average, this was only a 14% increase from the average for the previous week and only a 2% increase from share volume on the previous trading session of Friday, October 24. Similarly, while the 906.4 million shares traded on Nasdaq on October 27 represented a 42% increase over the year-to-date average and a 13% increase from the share volume average for the previous week, Nasdaq volume on October 27 was actually 2% less than on October 24.135

The statistics developed by OEA show the increase in the velocity of the stock market decline in the period leading up to the early close at 3:30 p.m. Specifically, the rate of the price decline in the S&P 500 index increased from 0.01% per minute in the interval from 9:30 a.m. to 1:03 p.m. to a rate of 0.03% per minute from 1:03 p.m. to 2:35 p.m., and the rate of decline increased even further, to 0.10%, in the interval from 3:05 p.m. to 3:30 p.m.136

Quote spreads widened as the market decline accelerated. The quote spreads for NYSE and Nasdaq stocks from 3:05 p.m. to 3:30 p.m. on October 27 were significantly wider than those during a comparable period on the control date of October 23. The quote spreads had also widened from those evident in the intervals earlier on October 27.137 The statistics developed by the NASD staff are comparable. Nasdaq spreads widened from 11% above normal prior to the first circuit breaker halt to levels 17% to 19% above normal for the interval 3:05 p.m. to 3:30 p.m. The statistics for the effective spreads for NYSE stocks from 3:05 p.m. to 3:30 p.m. on October 27 follow the same pattern.138

The OEA statistics on market liquidity and depth during this period, however, are mixed. For example, OEA found that the statistics for "bid down tick" trades and volumes for S&P 500 index and DJIA stocks generally showed significantly more depth in the interval from 3:05 p.m. to 3:30 p.m. on October 27 in comparison with the statistics for the earlier part of the day or those for the late afternoon on the October 23 control date.139 On the other hand, OEA's calculations regarding the average bid and ask depths in NYSE stocks from 3:05 p.m. to 3:30 p.m. on October 27 show that, overall, the increases in the ask sizes outpaced the increases in bid sizes in S&P 500 index and DJIA stocks during this period.140

In terms of NYSE specialist activity, OEA found that specialists in aggregate continued to increase their net buying in both DJIA and S&P 500 index stocks as the share price declines accelerated.141 The NASD staff reported that, in the final 24 minutes of trading on October 27, Nasdaq market makers bought $1.39 billion of stock and sold $1.31 billion. In the aggregate, market makers supported the market by buying $78 million dollars worth of stock from investors.

The price trends leading up to the 3:30 p.m. circuit breaker halt are generally consistent with the information provided to the Division in interviews with the most active trading firms and mutual fund complexes. Firms indicated that the market reopenings after 3:0