The quote exception rule: giving high frequency traders an unintended advantage.
"By the time the ordinary investor sees a quote, it's like looking at a star that burned out 50,000 years ago." Sal Arnuk, Partner, Themis Trading (Adler, 2012).
High frequency traders (HFTs) use fast computer algorithms to open and then close trading positions within milliseconds. A number of studies examine HFTs and conclude that HFTs benefit markets by providing liquidity and dampening volatility. (1) In contrast, we examine a specific case in which the structural problems created by latency and the trading of individual assets in multiple markets simultaneously, and attempts by regulators to level the playing field across these markets, may have created the unintended consequence of allowing fast trades to gain revenue from trading with slow traders. Specifically, we use data from US equity markets in 2008 to investigate the US Securities and Exchange Commission's (SEC) "Flickering Quote Exception" to the Regulation National Market System (Reg NMS) Order Protection Rule that defines the benchmark price for evaluation of trade throughs and provides a one second look back exception. (2)
For heavily traded companies, at least one exchange is always at the market wide best bid (National Best Bid and Offer [NBBO] Bid) and the market wide best ask (NBBO Ask) (definitions of terms are presented in Table I). Thus, the Benchmark Quote Exception implicitly defines the market wide reference price for the evaluation of a trade through as the least aggressive NBBO Ask and NBBO Bid (together the NBBO) over the previous one second of trading. We call these reference prices the Benchmark Ask and the Benchmark Bid, respectively, or taken together the Benchmark Quote. Prices equal to or better than the Benchmark Quote, but inferior to the NBBO Quote, are called Benchmark Compliant Prices. Any trades that execute at Benchmark Compliant Prices are not trade throughs under Rule 611. However, prices inferior to the Benchmark Quote may be accessed with the use of an Intermarket Sweep Order (ISO) (Chakravarty et al., 2012).
We document several behaviors of liquidity suppliers that indicate an active strategy to exploit the Benchmark Quote Exception. We find that as the distance in cents between the NBBO Ask (Bid) and Benchmark Ask (Bid) increases, relatively more exchanges quote at Benchmark Compliant Prices. When the distance is $.01,2.26 exchanges offer liquidity at NBBO prices, but only 1.73 exchanges offer liquidity at Benchmark Compliant Prices. When the gap is $.03, 1.68 exchanges offer liquidity at NBBO prices, but 2.19 exchanges offer liquidity at Benchmark Compliant prices.
In addition, Benchmark Compliant quotes are found on all exchanges, but are more prevalent on the smaller exchanges. Given a price change that initiates a Benchmark Event, the National Stock Exchange offers liquidity at Benchmark Compliant Prices 60% of the time and NBBO prices 40% of the time. The NYSE offers liquidity at NBBO prices 46% of the time and Benchmark Compliant Prices 53% of the time. For the NASDAQ, liquidity is offered at NBBO prices roughly 60% of the time and at Benchmark Compliant Prices 40% of the time. Substantial volume is executed at Benchmark Compliant Prices. For our sample of 100 thick stocks, we find that there are 107,878 trades per stock at Benchmark Compliant Prices on a typical stock day. For the full market, roughly 62 million shares trade at Benchmark Compliant prices on a day. We estimate that the Benchmark Quote Exception may have allowed HFTs to earn risk-free revenue of $233 million in 2008.
On March 10, 2008, the NYSE implemented a system upgrade that significantly decreased its latency by roughly 600 milliseconds. We use this event to examine how this speed change affected behavior during Benchmark Events. We find that for trades executed at NBBO prices, order execution quality improves significantly. For trades at Benchmark Compliant Prices, there is little or no improvement is order execution quality. However, we find a decrease in the average daily volume executed at Benchmark Compliant Prices indicating that faster markets reduce the ability of fast liquidity suppliers to execute the quote-arbitrage strategy.
Our findings are particularly timely as a number of regulatory proposals to police high frequency and algorithmic trading currently are under consideration. For example, Hong Kong's Securities and Futures Commission is currently proposing new regulations for algorithmic trading systems. (3) The Australian Securities and Investments Commission is recommending new rules to tighten controls over automatic trading. (4) In addition, Germany is aggressively moving to limit high frequency trading. (5) In the United States, the Financial Stability Oversight Council, comprised of the country's top financial regulators, is increasingly concerned with the effects of HFTs on other investors. (6) We believe that our findings on the impact of speed differentials in trading will be useful in these regulatory efforts.
Our research efforts are most closely aligned with Garvey and Wu (2010) who examine the impact of slow trading caused by the geographical distance between the trade initiator and the market center. They find that not only do traders located in close proximity to New York have lower execution costs and better execution quality, but that these traders trade differently, submitting and canceling more orders than their long distance counterparts. Our analysis differs in that we focus on how fast liquidity suppliers adjust quote strategies in a market with both fast and slow liquidity demanders. Hasbrouck and Saar (2012) develop a methodology for measuring the level of high frequency trading using widely available data and conclude that increased low latency trading improves market quality by lowering short-term volatility, decreasing spreads, and increasing displayed depth in the limit order book. Wissner-Gross and Freer (2010) determine that optimal locations exist from which to coordinate the statistical arbitrage of pairs of securities traded on two exchanges and identify the optimal location between the two exchanges. This location is often on an ocean floor. Budimir and Schweickert (2009) find that trading activity, time of day, and distance are the main drivers of latency.
The rest of the paper is organized as follows. Section I provides an overview of the SEC Reg NMS Order Protection Rule and the Benchmark Quote Exception and describes the quote-arbitrage trading strategy. Section II describes the data and methodology. Section III provides evidence regarding the existence and extent of arbitrage opportunities for fast liquidity suppliers. Section IV supplies evidence consistent with HFTs' strategic choice of order placements to increase the opportunity for quote-arbitrage, while Section V provides our conclusions.
I. Regulatory Background and Quote-Arbitrage Strategy
A. Reg NMS Order Protection Rule and Flickering Quote Exception
Figure 1 reports the three Benchmark Event states (States 1 3). The upper left panel illustrates the Ask-Improved-State. In this state, the NBBO Ask has declined, while the NBBO Bid has either declined (solid line) or remained unchanged (dotted line). The Benchmark Ask is greater than the NBBO Ask, while the Benchmark Bid is equal to the NBBO Bid. Any trade at a price at or below the Benchmark Ask, but above the NBBO Ask, is not a trade through. In a downward trending market, liquidity suppliers are able to sell at prices greater than the NBBO Ask.
Hasbrouck and Sarr (2009) investigate quotes that are quickly canceled, referring to these as fleeting orders. They indicate that motivations for fleeting orders include searching for hidden liquidity or enticing trades from counterparties who may not be willing to submit limit orders themselves. A fleeting quote, like all other quotes, does not initiate a Benchmark Event unless it improves the NBBO. Our research focuses on all quotes that improve the NBBO, which may include fleeting orders.
The lower left panel presents the Bid-Improved-State. In this state, the NBBO Bid has increased, while the NBBO Ask has either increased (solid line) or remained unchanged (dotted line). The Benchmark Bid is less than the NBBO Bid, while the Benchmark Ask is equal to the NBBO Ask. Any trade at a price at or above the Benchmark Bid, but below the NBBO Bid, is not a trade through. In an upward trending market, liquidity suppliers are able to buy at prices lower than the NBBO Bid.
The upper right panel provides the Two-Sided-State. Here, both the NBBO Ask and the NBBO Bid have narrowed over the previous one second, resulting in the ability of liquidity suppliers to trade at strictly better Benchmark Compliant Prices on both sides of the market. Typically, the Two-Sided-State state begins as either an Ask-Improved-State or a Bid-Improved-State.
The lower right panel reports the NBBO-State. Here, the NBBO Quote has remained unchanged (solid line) or one or both of the NBBO Ask and Bid have widened (dotted lines), so that the least aggressive NBBO Ask is the current NBBO Ask and the least aggressive NBBO Bid is the current NBBO Bid.
B. Quote-Arbitrage Strategy Used by Fast Liquidity Suppliers
To better integrate the various exchanges trading equities and to encourage the display of liquidity, the SEC adopted the Order Protection Rule, Rule 611, in 2005. Simply, the Order Protection Rule protects the best bid and ask at one exchange from trades at inferior prices at another exchange. However, the SEC recognized that in computer-driven markets, quotes within an exchange may update faster than the exchange can disseminate its new prices to other exchanges for the evaluation of an NBBO. The SEC feared that using the instantaneous quoted price from an exchange as the reference price for the assessment of trade through violations of the Order Protection Rule would create a myriad of trade through claims due to inter-market transmission latencies. Consequently, the SEC adopted the Benchmark Quote Exception, set forth in paragraph (b)(8) of Rule 611. This exception states that the reference prices for the evaluation of a trade through are the least aggressive ask and bid quotes over the previous one second from the exchange claiming the trade through.
The Order Protection Rule and the Benchmark Quote Exception establish the regulatory environment that forms the basis of our analysis. The markets have changed significantly since these rules were adopted. Intermarket communications have migrated from the high latency Intermarket
Trading System (ITS) to the low latency communications linkages that exist today and that are required by Reg NMS. High speed co-located computer systems, low latency intermarket communication linkages, and sophisticated trading algorithms for supplying liquidity to markets give centrally located fast liquidity suppliers (liquidity suppliers) an advantage over slow traders when prices change.
Suppose the NBBO Ask is 20.05 and updates to 20.03. The fast liquidity supplier can see this price change and if she has an outstanding limit order at Exchange 2 (EX2) at 20.05, while Exchange 1 (EX1) is displaying the best price of 20.03, she can choose not to cancel and update her quote to match the new best price. From the slow trader's time-delayed viewpoint, the best price in the market is still 20.05. If the slow trader submits a marketable buy order to EX2 that arrives within the current one second window of the Benchmark Quote Exception, the fast trader's limit order at 20.05 is executed without triggering a trade through. She can then quickly buy the security at EX1 at 20.03, closing out the arbitrage and netting $.02 per share. While this strategy requires a rapid pace, Hasbrouck and Saar (2012) find that liquidity suppliers can respond to changes in quotes on NASDAQ in two to three milliseconds. If an exchange is not posting the best price or not posting a price at all, the liquidity supplier does not have the option of submitting a limit order to that exchange at the Benchmark Price of 20.05 in an effort to pick off a slow trader. This is because slow traders observe quotes at a time lag, and will not observe the new price from the exchange until a later point in time when the slow trader will also observe the better price at the updating exchange.
Although higher revenue is earned by liquidity suppliers when trades are executed at Benchmark Prices, there is a lower probability of order execution relative to a quote at NBBO prices. This economic fact leads to a number of empirical implications. First, given the differences among exchanges in execution speed, trading costs, and order types, we believe that it is likely that fast traders will employ different strategies on different exchanges. The strategies of the fast traders will also depend upon the distribution of slow traders across exchanges. Hence, the liquidity suppliers on some exchanges will be more likely to update quotes to match any changes in the NBBO, while the liquidity suppliers on other exchanges will be more likely to compete at Benchmark Compliant Prices. Additionally, when the distance, in cents, from the NBBO price to the Benchmark Price increases, fewer exchanges are likely to quote at the NBBO price, seeking gains from the increased spreads available at Benchmark prices.
II. Data, Methods, and Sample
A. Data and Sample
Our sample period is from January 2, 2008 to August 29, 2008, which is 168 trading days. Our data are from the Daily Trade and Quote (DTAQ) data set, which unlike the Monthly Trade and Quote (MTAQ) data set used extensively in microstructure studies, is time stamped to the millisecond and also contains the exchange calculated NBBO Quote. Our sample period matches the time period of the DTAQ data set that we have. Our data set includes nine trading venues, each identified by its DTAQ code. Hereafter, for simplicity, we refer to these as exchanges although the Automated Display Facility is not an exchange.
For our primary sample, we select the 100 largest NYSE listed common stocks based on market capitalization on January 2, 2008. We exclude financial stocks (Standard Industrial Classification (SIC) code 6000) from our sample as the SEC banned naked short selling on selected financial stocks in July 2008. Because of this ban, during part of our sample period, trading in these stocks was unique so that inclusion of these stocks may make our results less general. Also, we restrict our sample to NYSE listed securities. The NASDAQ market provides quotes for NYSE listed firms, but the NYSE will not quote on NASDAQ firms. Consequently, NASDAQ firms have one less exchange offering liquidity, which would complicate our analysis without adding any significant benefit. We focus primarily on large firms since at least one exchange is always at the NBBO Bid and NBBO Ask.
B. Preferencing Measure
We cannot compare the spreads of trades executed at Benchmark Compliant Prices and NBBO Quote directly as high Benchmark Quotes result in higher effective spreads for the trades at Benchmark Compliant Prices. This issue is addressed by He, Odders-White, and Ready (2006) who propose the use of a Preferencing Measure (PM) that is defined as the ratio of realized spreads to effective spreads. We assess the execution quality of trades using PM. The lower the PM values for a trade, the better the execution quality.
C. Aligning Trades and Quotes
The DTAQ data set provides trades and quotes in separate files. A critical technical requirement of our analysis is to integrate the trades with the prevailing NBBO Quote at the time of the trade execution. We align trades and quotes as follows. First, we note that our analysis is focused on trades that are executed on an exchange rather than trades that are merely reported through an exchange. Trades that are executed off the exchange (including dark pool trades, internalized trades, and ECN trades) are reported to the consolidated tape indirectly through a Trade Reporting Facility (TRF). TRF trades are dropped from our analysis as we do not know the quote that was in force at the nonexchange executing venue and because the latency required to report the trade to the exchange introduces an unknown time shift.
Our alignment process makes the following assumption. Given that almost all trades are executed in the computerized matching engine of the exchange without human interaction or input, the most correct adjustment to align the NBBO Quote with trade prices is the time lag that maximizes the number of trades that execute at the NBBO Quote. Specifically, for each stock day in the sample, we test quote lag times from 0 to 1,500 milliseconds in 25 millisecond increments. The 25 millisecond (0.025 second) time step is selected as a compromise between computational time requirements and trade quote alignments.
We recognize that trading intensity may change throughout the trading day, possibly impacting the latency of the Securities Industry Automation Corporation (SIAC) computer system that generates the DTAQ database. Therefore, we also condition the lag value for each 30 minute segment of the trading day. In addition, the physical and wired distance between each exchange and the SIAC system will vary. To control for this variability, a separate lag adjustment is made for each exchange. We then select the quote lag that maximizes the number of trades that execute at the NBBO Quote for each stock day, segment, and exchange. In other words, for each day, stock, segment, and executing exchange, there is an individual lag adjustment.
The dotted line in Figure 2 reports the percentage of trades that executed at the NBBO using our maximization rule. Some of the remaining trades executed at prices superior than the NBBO are possibly due to hidden orders or price improvement. These are added to the NBBO trades as the dashed line in Figure 2. The remaining trades execute at prices inferior to the NBBO. These are comprised of trades at Benchmark Compliant Prices, and perhaps additional ISO trades and trade throughs. While the average lag time starts at about 625 milliseconds, on March 10, 2008 the average lag times drop substantially to only 75 milliseconds. This drop in lag times results from a major system upgrade for the NYSE and SIAC.
Figure 2 also indicates that within system latencies dropped by about 550 milliseconds. Unfortunately, we do not have the data to estimate the change to out of system latencies that would vary by participant location and capabilities and represent the change in the speed at which participants receive quote and trade information. SIAC documents indicate that the system upgrade more than doubled the ability to process trades and quotes. This processing assessment is based on the increase in message transmission rates and bandwidth increases from the August 30, 2007 to May 14, 2008 SIAC capacity reports. We believe that the system upgrade also significantly reduced quote latencies for fast traders, while having little impact on slow traders. We will investigate this topic in a later section.
We present similar results to those in Figure 2 in Table II. However, to conserve space, we confine the discussion to the intraday results. The mean lag represents the average number of milliseconds of lag between the in force NBBO Quote and the trade time stamp. This value is highest in the first and 13th segment, which are typically the periods of highest trading volume. % Trades at Quote represents the percentage of trades executing at the NBBO Quote for the maximizing lag (ML) and for lag times of ML [+ or -] 25 milliseconds. These results confirm that the maximizing area is relatively flat, with small changes of 1% to 2% of trades between the ML and the ML [+ or -] 25 lags. The percentage of trades at the NBBO Quote increases monotonically over the day. We also report the first and third quartiles of the lag values before and after the speed change of the NYSE demonstrating that not only did the speed change reduce latencies, it also reduced the variability of lag times.
III. Opportunities for Quote-Arbitrage and Analysis of Benchmark Events
A. Extent of Trading During Benchmark Events
If there is no or limited trading at Benchmark Compliant Prices, the potential economic gains from transacting at Benchmark Compliant Prices will be minimal. Therefore, in Table III, we report the Benchmark Compliant Volume and the NBBO Volume. For a typical stock, about 107,000 shares are traded per day at Benchmark Compliant Prices. We believe that this represents a substantial amount of trading sufficient to make Benchmark Events a significant economic event. Benchmark Compliant Volume conditioned on a Benchmark Event represents [107,878/(107,878 + 1,273,773)] = about 8% of total volume during Benchmark Events giving us an estimate of the proportion of slow traders in the market.
Table IV presents the average time that each possible state is in effect. The average time that a Bid-Improved-State or Ask-Improved-State prevails is about 7.25% each. Two-Sided-States have a much lower average percentage of time (1.45%). Overall, there is a Benchmark Event in effect for almost 16% of the trading day. We also report the standard deviation, median, and 25th and 75th percentiles of the distribution of Benchmark Event time percentages. These results indicate that there is substantial variability in the amount of time, on a daily basis, that a Benchmark Event prevails. We note that states where the NBBO Quote is locked or crossed are also included in our analysis. These markets are no different than regular markets in regard to Benchmark Events. Slow traders still see past quotes and cannot observe the better prices in the market. Locked and crossed markets are evaluated in Shkilko, Van Ness, and Van Ness (2008). Our results demonstrate that liquidity suppliers can potentially gain rents from slow traders over a significant part of the typical trading day.
B. Effect of NYSE Latency Reduction on Benchmark Events
All traders have some level of latency in observing quote updates. The maximum possible speed that any trader can achieve is limited by the speed of the exchange. If the exchange is relatively slow at disseminating quotations, even fast traders might trade at Benchmark Quotes. If the market increases the speed of quote dissemination, fast liquidity demanders and suppliers are better able to take advantage of the reduced latency to target liquidity at NBBO Quotes.
On March 10, 2008, the NYSE migrated to a new, more powerful platform that significantly increased computational and communication speed. Using March 10 as the event day, we compare the effective half spread, the realized half spread, and the PM of slow trades trading at Benchmark Compliant Prices and fast traders trading at NBBO Quotes and present the results in Table V. Using a mean difference test, for trades at the NBBO, we reject the hypothesis of equality in each case indicating that faster markets are beneficial to fast traders. These results are similar to those of Riordan and Storkenmaier (2012) for the Deutsche Boerse. However, for trades at Benchmark Compliant Prices, we reject the hypothesis of equality for realized spread and PM. We cannot reject the hypothesis of equality for the effective spread.
The improvement in execution quality for slow traders is driven mainly by a reduction in realized spreads. We also examine the volume transacted during Benchmark Events at Benchmark Compliant Prices and NBBO prices. We find that there is a significant reduction in both Benchmark and NBBO volume after the speed increase of the NYSE. The ratio of Benchmark Compliant Volume to NBBO Volume decreases from 0.093 to 0.080, indicating that the proportion of slow traders decreases after the speed increase. We believe that faster markets allow the marginally fast liquidity demanders to better target best prices, reducing Benchmark Compliant Volume.
To extend our analysis and control for variables that could possibly affect execution quality for fast and slow traders in the pre- and postperiods, we estimate the following equation using ordinary least square:
[PM.sub.i,t] = [alpha] + [[beta].sub.1]Spd + [[beta].sub.2]LnTrd[Cnt.sub.i,t] + [[beta].sub.3]Mp[Var.sub.i,t] + [[beta].sub.4]AvgTrd[Sz.sub.i,t] + [[beta].sub.5]Tm[Flk.sub.i,t] + [[epsilon].sub.t], (1)
where PM is the Preferencing Measure. Spd is a dummy variable that is equal to zero prior to March 10, 2008 and one otherwise. LnTrdCnt is the log of the number of trades executed, MpVar is the NBBO Quote midpoint variance for the full trading day, AvgTrdSz is the average trade size, and TmFlk is the percentage of the trading day with Benchmark Events. We estimate the regression separately for trades at Benchmark Compliant Prices and the NBBO Quote. When we estimate the regression by stock, we test whether the average coefficient is statistically different from zero.
Table VI reports our regression results. The key result is that the speed dummy variable is insignificant for trades at Benchmark Compliant Prices, but significantly negative for trades at the NBBO Quote. This result confirms the univariate analysis that the decrease in latency on the NYSE improves the execution quality for fast traders. We believe that as market speed increases, fast traders are better able to target executions at NBBO Quotes improving their execution quality. Faster market speed does not help slow liquidity demanders who still observe the market in a past state.
We replicate the regression analysis using Effective Spread, and Realized Spread, in turn, as the dependent variable, but we do not present the results because of their similarity to those presented here.
C. Revenue Associated with Benchmark Events
Our results indicate that strategic liquidity suppliers can earn substantial revenue by posting liquidity at Benchmark Compliant Prices resulting in considerable economic benefits to liquidity suppliers. Numerically, the average Benchmark Compliant Volume per stock day is 107,878 shares at an average increase in effective spreads of about 1.3 cents (from Table VI, Row 1; 1.919 - 0.611 [approximately equal to] 1.3). With 100 companies and 168 days in our sample, the increased revenue to liquidity suppliers trading at Benchmark Compliant Prices is (100 companies x 168 days x 1.3 cents per share x 107,878 shares per day) = $23.6 million dollars. Projected over a typical 252 day trading year, the revenue is estimated at $35.4 million. Since the Benchmark Quote Exception is applicable to all exchange traded equities, such as Exchange Traded Funds, Real Estate Investment Trusts, and preferred stock, we estimate the revenue impact for an extended sample of equities next.
Next, we evaluate the impact of Benchmark Events for all securities (common stocks, Exchange Traded Funds, Real Estate Investment Trusts, and preferred stock) in the DTAQ database, from all listing exchanges, with a closing price over $5 each day for the week of March 24, 2008. This period is two weeks after the platform migration of the NYSE, allowing time for the market to stabilize. First, we align trades prices and (NBBO) quotes based on the method outlined previously. Then, for Benchmark Events, we evaluate the Benchmark Compliant Volume and the additional revenue obtained by liquidity suppliers by offering liquidity at Benchmark Compliant Prices.
Table VII provides the results of our market-wide analysis. We report the mean daily value for each variable. From a policy standpoint, we feel that the one second definition of the Benchmark Quote Exception is excessive in today's market. We estimate the impact of reducing the Benchmark Quote look back time by 200 millisecond increments. These values are located in the column labeled Benchmark Event Time. We calculate the additional revenue to fast liquidity suppliers from trades at Benchmark Compliant Prices. If the execution price is greater than the NBBO Ask, we compute the revenue per share as the execution price minus the NBBO Ask. If the execution price is less than the NBBO Bid, we compute the revenue per share as the NBBO Bid minus the execution price. The total revenue per trade is the revenue per share multiplied by the number of shares in the transaction. As the Benchmark Quote look back time is decreased, we drop trades that occur at prices inferior to the Benchmark Quote as these trades would be rerouted to exchanges quoting at the NBBO Quote and would not be executed at the higher price contained in the DTAQ record.
Under the current Benchmark Quote Exception's 1,000 millisecond look back for the Benchmark Quote, the average revenue to liquidity suppliers is about $0.93 million per day. In a typical year of 252 trading days, this yields $233 million of revenue. In addition, our results indicate that over 62 million shares are traded at Benchmark Compliant Prices. As the Benchmark Quote look back time is reduced from 1,000 milliseconds to 200 milliseconds, liquidity suppliers' revenue drops to $0.51 million per day or $129 million per year. This analysis is conservative as we are not able to take into account the decreased incentive to quote at Benchmark Compliant Prices that would result from a reduction in the allowed look back time. In addition, since TRF trades are not included in our analysis due to potential alignment issues with trades and NBBO Quotes, there is likely additional revenue that we are not capturing.
While we evaluate Benchmark Quote look back times as low as 200 milliseconds, we are not specifically recommending that the Benchmark Quote look back time should be reduced to 200 milliseconds. Instead, we feel that the Benchmark Quote look back time should be set based on current market conditions that evaluate the inter-exchange latency required to disseminate the NBBO Quote. This look back time could be well under 200 milliseconds in current markets. Overall, these results indicate that HFTs' revenue from taking advantage of the Benchmark Quote Exception is economically significant, highly profitable for fast traders, and very costly for slow traders.
IV. Strategic Choice of Order Placements to Increase the Opportunity of Quote Arbitrage
A. Regulatory Evaluation
One potential cause of our current and proceeding results is that the Order Protection Rule is simply ignored by the exchanges. We investigate this possibility and present our results in Table VIII. Specifically, we evaluate the median volume that executes at prices inferior to the NBBO during Benchmark Events and in the 1,000 milliseconds after a Benchmark Event has concluded. Panel A reports the volume traded during Benchmark Events at Benchmark compliant prices and NBBO prices, conditioned on trade type, ISO and Non-ISO (NISO). Panel B presents the trading volume in the post-Benchmark Event. Our results indicate that little volume trades outside of NBBO prices in the post-Benchmark Event period, regardless as to whether the trades are ISO or NISO. This evidence indicates that the Order Protection Rule is closely followed over the time period of our analysis.
B. Competitive Focus Analysis
One implication of the potential of earning economic gains from trading at Benchmark Compliant Prices is that these liquidity suppliers forgo trading with any liquidity demanders that see the current state &the market. These fast traders see the better prices in the market and, therefore, route to those exchanges. We investigate whether an exchanges' competitive focus varies relative to the propensity to quote at Benchmark Compliant Prices. The results are shown in Table IX. We arrange the exchanges in order (from the lowest in Column 2 to the highest in Column 9) by percentage of time quoting at the NBBO Quote in the Bid-Improved-State, normalized to 100%. Specifically, we report the value of [(Time at NBBO Quote/(Time at NBBO Quote + Time at Benchmark Compliant Price)) x 100] and its complement. We omit the CBOE due to its very small market share. We reject the null hypothesis of equality of means at the 0.01 level. Thus, we conclude that exchanges differ in their competitive focus with regard to quoting at the NBBO Quote and at Benchmark Compliant Prices.
To investigate exchanges' competitive focus further, for the 91,833,932 Benchmark Events during our sample period, we calculate the percentage initiated by each exchange. In Table X, we present the percentages arranged from the lowest percentage in the Bid-Improved-State in Column 2 to the highest in Column 10. We reject the hypothesis of equality of means. Just four exchange, International Stock Exchange, NASDAQ, NYSE, and ARCA/Pacific, initiate 97.5% of the Benchmark Events. A higher level of initiation of Benchmark Events indicates a competitive focus on establishing the best prices in the market.
Note that we do not explore why or how each exchange achieves its competitive focus. Exchanges compete on a number of dimensions including fees, order types, and execution speed. For our purposes, it is sufficient that the differing portfolios of services result in differences in the attractiveness of competing at NBBO Prices versus Benchmark Compliant Prices.
C. Benchmark Quote Competition as a Function of the Range of Benchmark Compliant Prices
Next, we investigate whether a larger Max Gap entices more exchanges to quote Benchmark Compliant Prices. The larger the price range between the NBBO and Benchmark Price, the greater the potential revenue from not updating prices to match the NBBO. For each Max Gap, in Table XI, we present the time weighted average number of exchanges quoting at the NBBO and at Benchmark Compliant Prices. We refer to the number of exchanges quoting at a specified price as the breadth of the market. Recall that all of the observations occur during Benchmark Events. The Pearson correlation coefficient between the Max Gap and the percent of exchanges quoting at NBBO prices (%NBBO Breadth) is -0.360, which is statistically significant at the 0.01 level indicating that when the Max Gap is larger, the number of exchanges quoting at NBBO prices is lower. This provides strong evidence that as the Max Gap increases, exchanges are drawn away from quoting at the NBBO Quote and toward quoting at Benchmark Compliant Prices.
Table XI provides the number of Benchmark Events at each Max Gap, including only those Benchmark Events initiated with an increase in the Max Gap. This basically captures the number of instances in which a given stock's price has increased (or decreased) by $.01 through $. 10+ cents during the last second. The number of Benchmark Events at each successively higher Max Gap decreases rapidly.
Table XI, Panel B, which limits our observations to those Benchmark Events with a life of at least 50 milliseconds, sheds additional light on exchange quoting behavior. Most Benchmark Events exceed 50 milliseconds and the proportion that do increases as the Max Gap increases. The strong downward trend in Column 5 is still evident here. The coefficient of correlation between Max Gap and the %NBBO Breadth, which is -0.370, is statistically significant at the 0.01 level.
The speed of entry and either execution or cancellation of quotes has increased dramatically. However, not all traders have kept pace in their ability to deal with this fast paced trading environment. In a market with both slow and fast traders, we investigate how fast strategic liquidity suppliers can submit limit orders with a high probability of transacting against slow traders. The SEC's Benchmark Quote Exception to the Order Protection Rule allows trades on an exchange to occur at prices inferior to the best contemporaneous prices on other exchanges as long as the trade occurs at a price equal to or better than the Benchmark Quote. The Benchmark Quote is the least aggressive NBBO Bid and the least aggressive NBBO Ask over the past one second. Benchmark Compliant Prices are prices equal to or better than the Benchmark Quote, but inferior to the NBBO. Since fast traders see the market prices in real time, they only route trades to markets with quotes at the NBBO Quote. Slow traders observe quotes with a time delay so they may not observe a recent quote update. Consequently, they may route orders to markets with quotes at prices inferior to the NBBO Quote. These orders may be executed at Benchmark Compliant Prices under SEC Rule 611.
We investigate this market environment and obtain the following results. The propensity to quote at the NBBO Quote and at Benchmark Compliant Prices varies across exchanges. The International Stock Exchange, NYSE, ARCA/Pacific, and NASDAQ quote most often at the NBBO Quote, while the Chicago Stock Exchange, the National Stock Exchange, the Automated Display Facility, the Philadelphia Stock Exchange, and the CBOE are more likely to quote at Benchmark Compliant Prices. Furthermore, the level of competition at the NBBO Quote decreases as the gap between the Benchmark Compliant Price and the NBBO Quote increases.
Based on our sample of 100 NYSE stocks and 168 days of trading, we estimate that liquidity suppliers that strategically quote at Benchmark Compliant Prices earn $35.4 million in a typical year. Using an extended sample of all equities in the DTAQ database with a price over $5, we estimate that the market wide impact of the Benchmark Quote Exception is $233 million per year in revenue to fast liquidity suppliers at the expense of slow traders.
On March 20, 2008, the NYSE migrated to a significantly faster computer platform. As a result of this migration, latencies dropped more than 600 milliseconds. We use this event to examine how increases in market speed impact fast and slow traders in the context of the Benchmark Quote Exception. Our results indicate that execution quality for fast traders improves significantly as market speed increases. However, execution quality for slow traders improves only slightly. Our analysis confirms that as markets become faster, fast liquidity demanders are better able to avoid trading at Benchmark Quotes. However, fast liquidity suppliers are also better able to target slow traders at Benchmark Quotes, earning rents. We find that while both effective and realized spreads of trades at the NBBO Quote decrease after the market speed increases, the execution quality of trades at Benchmark Quotes remains essentially unchanged.
As global markets become more integrated and communication speeds decrease, regulators will face additional challenges in addressing the speed differentials of market participants both inside and outside of their regulatory jurisdiction. If a stock trades on both the New York and London exchanges, a strategically placed HFT system on the floor of the Atlantic Ocean will be able to execute the quote arbitrage strategy, but is outside of the jurisdiction of both countries regulators. Greater intergovernmental regulatory cooperation will be required to address these issues.
Adler, J., 2012, "Raging Bulls: How Wall Street Got Addicted to Light-Speed Trading," Wired (magazine). August 3. Downloaded 12 September 2012 from http://www.wired.com/business/2012/08/ff_wallstreet_trading/all/.
Brogaard, J., 2010, "High Frequency Trading and Its Impact on Market Quality," Northwestern University Working Paper.
Budimir, M. and U. Schweickert, 2009, "Latency in Electronic Securities Trading: A Proposal for Systematic Measurement," Journal of Trading 4, 47-55.
Chakravarty, S., E Jain, J. Upson, and R. Wood, 2012, "Clean Sweep: Informed Trading Through Intermarket Sweep Orders," Journal of Financial and Quantitative Analysis 47, 415-435.
Garvey, R. and E Wu, 2010, "Speed, Distance, and Electronic Trading: New Evidence on Why Location Matters," Journal of Financial Markets 13,367-396.
Hasbrouck, J. and G. Saar, 2009, "Technology and Liquidity Provision: The Blurring of Traditional Definitions," Journal of Financial Markets 12, 143-172.
Hasbrouck, J. and G. Saar, 2012, "Low-Latency Trading," AFA 2012 Chicago Meetings Paper Johnson School Research Paper Series No. 35-2010, July 27.
He, C., E. Odders-White, and M. Ready, 2006, "The Impact of Preferencing on Execution Quality," Journal of Financial Markets 9, 246-273.
Hendershott, T. and R. Riordan, 2011, "Algorithmic Trading and Information," NET Institute Working Paper.
Riordan, R. and A. Storkenmaier, 2012, "Latency, Liquidity, and Price Discovery," Journal of Financial Markets 15, 416-437.
Shkilko, A., B. Van Ness, and R. Van Ness, 2008, "Locked and Crossed Markets on NASDAQ and the NYSE," Journal of Financial Markets 11,308-337.
Wissner-Gross, A.D. and C.E. Freer, 2010, "Relativistic Statistical Arbitrage," Physical Review E 82, 056104, 1-7.
The authors wish to thank an anonymous referee. Sie Ting Lau, Bohui Zhang, William T. Smith, Larry Harris, Joel Hasbrouck, Gideon Saar, Maureen O'Hara, and Jonathan Brogaard for helpful comments. We received useful comments from the participants at a Securities and Exchange Commission seminar in September 2011 and the NBER market microstructure conference in December 2011. We also thank Leslie Boni, Marc Lipson, Michael Goldstein, and other seminar participants at the 2012 FMA. Any errors remain our own.
(1) According to Brogaard (2010), HFTs participate in 77% of the dollar volume traded, generating $2.8 billion of gross annual trading profits. Moreover, since HFTs rarely trade with each other, these revenues come from non-HFTs. In return for incurring these costs, markets benefit as HFTs are frequently at the best quotes and dampen market volatility. Hasbrouck and Saar (2012) conclude that increased low latency activity by HFTs improves traditional market quality measures such as short-term volatility, spreads, and displayed depth. Hendershott and Riordan (2009) examine algorithmic traders, of which HFTs are a subset, and conclude that they consume liquidity when it is cheap, supply liquidity when it is expensive, and help move prices toward the efficient price. Garvey and Wu (2010) report that speed advantaged traders engage in strategies that are more conducive to speed.
(2) SEC release 34-51808 discusses the "Flickering Quotes Exception" on page 152. See www.sec.gov/rules/final/3451808.pdf. However, many understand the term flicker:.ng quote to refer to instances where a single exchange flickers the quote price to indicate additional depth outside of the best price on the exchange. Hereafter, we substitute the term "benchmark" for the term "flicker" to avoid potential confusion with flickering quotes.
Thomas H. McInish * and James Upson
* Thomas H. McInish is a Professor and Wunderlich Chair of Finance in the Department of Finance, Insurance and Real Estate at Fogelman College of Business and Economics at the University of Memphis in Memphis, TN. James Upson is an Assistant Professor of Finance in the College of Business at the University of Texas at El Paso in El Paso, TX.
Table I. Key Terms Term Definition General Terms Breadth The time weighted number of exchanges offering depth at a specified price. The price can be the NBBO Ask, the NBBO Bid, or a Benchmark Compliant Price. Benchmark Event A period when Benchmark Compliant Prices are in effect. Fleeting Order A quote that was intended to be executed immediately against hidden or unexpressed liquidity, but to be cancelled otherwise. Benchmark Quote An exception to the Order Protection Rule Exception that defines the Benchmark Quote for the evaluation of a trade through. As long as an exchange's trades are at prices equal to or better than the Benchmark Quotes, other exchanges cannot claim that an exchange is trading through their better quotes. Order Protection Rule Rule 611, adopted by the SEC in 2005. This rule protects top of the book limit orders from trade throughs at inferior prices on other exchanges when these limit orders are eligible for NBBO Quote participation. Preferencing Measure The ratio of realized spreads to effective spreads defined by He et al. (2006). Four Market States Ask-Improved-State The market state in which the Benchmark Ask is greater than the NBBO Ask. In this state, buyer-initiated trades can execute at prices above the NBBO Ask, but at or below the Benchmark Ask without violating the Order Protection Rule. Bid-Improved-State The market state in which the Benchmark Bid is less than the NBBO Bid. In this state, seller-initiated trades can execute at prices less than the NBBO Bid, but at or above the Benchmark Bid without violating the Order Protection Rule. Two-Sided-State The market state where the Benchmark Ask is greater than the NBBO Ask and the Benchmark Bid is less than the NBBO Bid. Trades can occur at Benchmark Compliant Prices that are inferior to the NBBO without violating the Order Protection Rule. NBBO-State In this state, the Benchmark Bid equals the NBBO Bid and the Benchmark Ask equals the NBBO Ask. This state occurs when: 1) the NBBO Quote has remained unchanged for one second, 2) both the NBBO Ask and NBBO Bid widened, or 3) either the NBBO Ask or the NBBO Bid has widened and the other side of the market has remained unchanged for one second. Benchmark Event Terms Benchmark Ask Depth The total depth on the ask side offered at Benchmark Compliant Prices excluding depth at the NBBO Ask. Benchmark Bid Depth The total depth offered on the bid side at Benchmark Compliant Prices excluding depth the NBBO Bid. Benchmark Compliant A price equal to or better than the Benchmark Price Ask or Benchmark Bid, or both, but inferior to the NBBO Quote. Benchmark Compliant The sum of the Benchmark Ask Depth and the Depth Benchmark Bid Depth. Benchmark Compliant A trade that occurs at a Benchmark Compliant Trade Price. Benchmark Compliant The total volume associated with Benchmark Volume Compliant Trades. Benchmark Spread The difference between the Benchmark Ask and the Benchmark Bid. Ask Gap The difference in cents between the NBBO Ask and a Benchmark Compliant Price on the ask side. Benchmark Ask The benchmark used to evaluate trade throughs on the ask side of the market, defined by the SEC as the least aggressive (highest) NBBO Ask over the previous one second of trading. If there has been no change in the NBBO Ask during the last one second, the Benchmark Ask equals the NBBO Ask. Benchmark Bid The benchmark used to evaluate trade throughs on the bid side of the market, defined by the SEC as the least aggressive (lowest) NBBO Bid over the previous one second of trading. If there has been no change in the NBBO Bid during the last one second, the Benchmark Bid equals the NBBO Bid. Benchmark Quote The Benchmark Ask and the Benchmark Bid taken together. Bid Gap The difference in cents between the NBBO Bid and a Benchmark Compliant Price on the bid side. Max Ask Gap The difference in cents between the NBBO Ask and the Benchmark Ask. If the Benchmark Ask is 20.05 and the NBBO Ask is 20.01, then the Max Ask Gap is .04. Max Bid Gap The difference in cents between the NBBO Bid and the Benchmark Bid. If the Benchmark Bid is 19.96 and the NBBO Bid is 20.00, then the Max Bid Gap is .04. NBBO Ask The market wide best ask. NBBO Ask Depth The aggregate depth offered at the NBBO Ask. NBBO Bid The market wide best bid. NBBO Bid Depth The aggregate depth offered at the NBBO Bid. NBBO Depth The sum of the NBBO Ask Depth and the NBBO Bid Depth. NBBO Volume The number of shares traded at the NBBO. NBBO Quote The NBBO Ask and the NBBO Bid together. Table II. Trade Quote Alignment by Exchange We present descriptive statistics on our alignment of trades and NBBO Quotes. For stock i on day t for each 30-minute segment of the trading day j for trades from exchange p, we test quote lags from 0 to 1,500 milliseconds in 25 millisecond increments, in turn. We then select the lag that maximizes the percentage of trades that execute at the NBBO Quote. Mean Lag (ML) is the mean optimal quote lag time in milliseconds. The NYSE implemented a significant computer system upgrade on March 10, 2008. Pre (Post) results are based on alignment results prior to (after) the upgrade. % Trades at Quote reports the average percentage of trades for the indicated segment. Columns 4, 5, and 6, respectively, present the percentage of trades at the quote for ML-25, ML, and ML+25. We also report the first and third quartiles of the maximizing lag for each stock day and stock day segment for the pre- and postperiods. Mean Lag (a) % Trades at Quote Pre Post ML-25 At ML ML+25 Full Day 719.4 94.2 84.89 85.29 84.88 Intraday 1 668.5 84.1 78.14 79.60 78.51 2 619.9 77.5 82.58 83.85 82.85 3 605.1 73.7 84.17 85.53 84.55 4 601.5 74.0 85.20 86.64 85.68 5 595.3 71.8 85.73 87.28 86.29 6 594.6 70.7 85.97 87.57 86.57 7 590.3 70.9 86.25 87.96 86.94 8 592.9 70.5 86.15 87.46 86.45 9 591.9 71.7 86.47 88.10 87.13 10 594.3 72.9 85.82 87.25 86.30 11 592.0 73.1 86.64 88.07 87.14 12 593.6 75.1 87.17 88.35 87.47 13 594.3 78.6 87.67 88.55 87.84 First and Third Quartiles (a) Pre Post Q1 Q3 Q1 Q3 Full Day 600 875 25 125 Intraday 1 550 850 25 100 2 450 825 0 100 3 400 850 0 100 4 375 850 0 100 5 325 850 0 100 6 325 850 0 100 7 300 850 0 100 8 325 850 0 100 9 300 850 0 100 10 300 850 0 100 11 300 850 0 100 12 300 850 0 100 13 300 850 0 100 (a) In milliseconds. Table III. Benchmark Compliant Volume and NBBO Volume We present the trade volume at Benchmark Compliant Prices and at the NBBO Quote for each state. Volume is the mean daily trading volume for each stock for each day during Benchmark Events, excluding TRF volume. Volume Benchmark Event NBBO Bid-Improved-State 37,186 516,087 Ask-Improved-State 36,516 515,297 Two-Sided-State Bid 17,299 121,544 Ask 16,877 120,845 Total 107,878 1,273,773 Table IV. Percentage of Time that the Market is in Each State As demonstrated in Figure 1, there are four possible states: 1) Ask-Improved-State, 2) Bid-Improved-State, 3) Two-Sided-State, and 4) NBBO-State. Total is the sum for the first three states. For the period January 2, 2008 to August 29, 2008, for the 100 largest NYSE listed common stocks (excluding financial firms), we present the percentage of time that the market is in each state. Bid-Improved- Ask-Improved- Two-Sided- Statistic State State State Mean 7.24% 7.25% 1.45% Std 4.47% 4.49% 1.83% Median 6.33% 6.32% 0.78% 25% 3.89% 3.88% 0.35% 75% 9.95% 9.94% 1.82% NBBO Statistic Total State Mean 15.93% 84.07% Std 10.53% 10.53% Median 13.50% 86.50% 25% 8.19% 78.20% 75% 21.80% 91.81% Table V. Cost and Market Quality Comparison We calculate the daily effective half spread, realized half spread, and the Preferencing Measure (PM) during Benchmark Events for each stock day and present the means for the period before (Pre) and after (Post) March 10, 2008. Using a mean difference test, we test the null hypothesis that there is no difference in Pre and Post-event means. We present the respective mean differences in Columns 4 and 7. Benchmark Trades Pre- Post- Mean Measure Event Event Difference Effective Spreads 1.998 1.919 -0.079 Realized Spreads 1.242 1.103 -0.139 * Preferencing Measure 0.745 0.730 -0.015 * Volume 159,249 88,447 70,802 * NBBO Trades Pre- Post- Mean Measure Event Event Difference Effective Spreads 0.684 0.611 -0.073 * Realized Spreads 0.309 0.185 -0.124 * Preferencing Measure 0.298 0.087 -0.210 * Volume 1,713,680 1,107,374 606,306 * * Significant at the 0.10 level. Table VI. Multivariate Analysis We estimate the following equation both as a fixed effects regression and at the stock level for trades at both Benchmark Compliant Prices and the NBBO: [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where PM (Preferencing Measure) is the daily average PM of trades at Benchmark Compliant Prices or at the NBBO, in turn, Spd is a dummy variable that is equal to one for days on or after March 20, 2008, and zero otherwise, LnTrdCnt is the log of the number of executed trades, Mp rar is the NBBO Quote mid-point volatility for Day t, AvgTrdSz is the average size of each trade, and TmFlk is the percentage of the trading day with Benchmark Events. At the stock level, we estimate the equation for each stock and then report the average coefficients. We test the null hypothesis that the mean of these coefficients equals zero. Benchmark Compliant Trades Cross Sectional Stock Level Intercept 0.907 *** 0.456 *** Spd -0.022 -0.034 LnTrdCnt 0.007 0.040 MpVar 0.024 ** 0.734 * AvgTrdSz 0.000 * 0.001 ** TmFlk -0.279 -1.082 * Effects Fixed R-Sq 0.010 N 16,760 Trades at NBBO Quote Cross Sectional Stock Level Intercept -1.099 *** -0.097 Spd -0.208 *** -0.211 *** LnTrdCnt 0.172 *** 0.070 *** MpVar 0.037 0.981 AvgTrdSz 0.000 ** 0.001 *** TmFlk -0.974 *** -1.539 *** Effects Fixed R-Sq 0.047 N 16,760 *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table VII. Estimated Market-Wide Benchmark Event Revenue For the week of March 24, 2008, we evaluate the market wide impact of the Benchmark Quote Exception and assess the sensitivity of Benchmark Compliant Volumes and revenue to the length of Benchmark Quote look-back time. As the Benchmark Quote look-back time decreases, trades that occur outside of the Benchmark Quote are dropped from the analysis since under the Order Protection Rule, these trades would be re-routed to exchanges posting better prices. Benchmark Compliant Volume is the total volume executed at Benchmark Compliant Prices. Benchmark Event revenue is calculated as the difference between the execution price and the NBBO Bid or NBBO Ask, as appropriate, multiplied by the number of shares traded, and summed for each day. Results presented are the daily average over the five trading days. On average, there are 7,090 observations per day. Benchmark Volume Revenue Event Time (Shares) (Dollars) 1,000 62,119,848 924,759 800 58,700,724 865,840 600 54,456,892 791,803 400 48,462,307 696,702 200 36,576,615 512,216 Table VIII. Benchmark Event Volume and Post-Event Volume We present the median daily volume of intermarket sweep order (ISO) and non-intermarket sweep order (NISO) trades over the maximum length of a Benchmark Event (Panel A) and for the 1,000 milliseconds of trading following the Benchmark Event (Panel B). If Benchmark Events are sequential, so that a second Benchmark Event is initiated prior to the 1,000 millisecond span of the first Benchmark Event, all trades at the Benchmark Compliant Prices of the first Benchmark Event are assigned to the first Benchmark Event over the full 1,000 milliseconds. Only trades that are outside of the first Benchmark Quote, yet at or inside the second set of Benchmark Compliant Prices are counted in the second Benchmark Event. Benchmark Events terminate either after the 1,000 milliseconds or when the NBBO-State returns. For the post-event evaluation, we require no Benchmark Events to occur over the full 1,000 milliseconds post-event period (i.e., the NBBO-State must prevail over the full 1,000 milliseconds). In the post-event analysis, a Benchmark Compliant Trade is defined as a trade at an inferior price to the NBBO, but at or inside of the posted quote for the executing exchange. Trades: ISO NISO ISO NISO Benchmark Volume NBBO Volume Panel A. During Benchmark Event (Milliseconds (a)) 0-50 1,300 2,300 14,657 22,895 51-200 2,600 5,100 24,796 43,460 201-400 2,527 4,100 21,300 44,500 401-600 2,100 3,000 18,000 34,075 601-800 2,100 2,900 18,300 33,252 801-1,000 3,200 4,600 28,304 49,900 Full 15,100 23,400 134,376 239,642 Panel B. Post-Event (Milliseconds (b)) 0-50 0 0 1,527 3,100 51-200 0 100 3,500 8,000 201-400 0 100 3,800 9,345 401-600 0 100 3,600 8,300 601-800 0 100 4,100 8,679 801-1,000 300 300 11,300 19,000 Full 800 1,200 29,100 57,861 (a) After initiation of Benchmark Event. (b) After end of Benchmark Event. Table IX. Exchange Competitive Focus We examine the competitive focus of exchanges during Benchmark Events. Beginning with the times when the market is in the Bid-Improved-State, for each exchange for each stock for each day, we calculate [(Time at NBBO Quote/(Time at NBBO Quote + Time at Benchmark Compliant Quote) x 100] and report the result in the first row labeled NBBO Quote in Panel A. The value reported in each column of the second row is 100% minus the value in the respective column of the first row. We present the results beginning with the exchange with the smallest value for NBBO Quote in the Bid-Improved-State and continuing with successively higher values. We repeat the analysis for the Ask-Improved-State and Two-Sided-States and present the results in Panels B and C, respectively. For Panels B and C, we retain the ordering of Panel A. The Ask and Bid sides of the market are combined for the Two-Sided-State. The F-statistic is from an equality of means test. Although not reported, we also conduct a Tukey and Bonferroni test for equality and obtain similar outcomes. BC = Benchmark Compliant. Exchange D M X C N Panel A. Bid-Improved-State NBBO 17.3% 26.6% 38.1% 39.9% 46.5% BC 82.7% 73.4% 61.9% 60.1% 53.5% Panel B. Ask-Improved-State NBBO 18.8% 36.4% 39.1% 43.5% 46.6% BC 81.2% 63.6% 60.9% 56.5% 53.4% Panel C. Two-Sided-State NBBO 21.0% 21.9% 16.5% 35.2% 48.8% BC 79.0% 78.1% 83.5% 64.8% 51.2% Exchange P T I F-statistic Panel A. Bid-Improved-State NBBO 57.8% 58.6% 62.5% 10,380 * BC 42.2% 41.4% 37.5% Panel B. Ask-Improved-State NBBO 57.8% 58.6% 62.7% 7,700 * BC 42.2% 41.4% 37.3% Panel C. Two-Sided-State NBBO 44.4% 44.7% 48.5% 8,362 * BC 55.6% 55.3% 51.5% Note: C = National Stock Exchange, D = Automated Display Facility, I = International Stock Exchange, M = Chicago Stock Exchange, N = NYSE, P = ARCA/Pacific, T = NASDAQ, X = Philadelphia Stock Exchange, and W = CBOE. * Significant at the 0.10 level. Table X. Initiation of Benchmark Events, by Exchange We report the percentage of Benchmark Events initiated on each exchange. Bid-Improved-State, Ask Improved-State, and Two-Sided-States are reported in Panels A, B, and C, respectively. For each state, we report results for the full sample period and for the period before (Pre) and after (Post) the NYSE speed change. F-Stat is the F statistic of a joint test of equality. X W M D C P Panel A. Bid-Improved-State Full 0.0% 0.0% 0.1% 0.1% 2.2% 16.8% Pre 0.0% 0.1% 0.0% 0.2% 1.8% 12.9% Post 0.0% 0.0% 0.1% 0.1% 2.4% 18.3% Panel B. Ask-Improved-State Full 0.0% 0.0% 0.1% 0.1% 2.3% 17.0% Pre 0.0% 0.1% 0.0% 0.1% 1.8% 12.9% Post 0.0% 0.0% 0.1% 0.1% 2.5% 18.5% Panel C. Two-Sided-State Full 0.0% 0.2% 0.2% 0.8% 3.2% 13.0% Pre 0.0% 0.4% 0.1% 0.6% 2.8% 11.5% Post 0.0% 0.0% 0.2% 0.9% 3.3% 13.5% N I T F-Stat Panel A. Bid-Improved-State Full 19.4% 30.6% 30.7% 42,571 * Pre 24.2% 30.5% 30.3% 11,734 * Post 17.6% 30.7% 30.8% 32,484 * Panel B. Ask-Improved-State Full 19.0% 30.9% 30.6% 43,228 * Pre 24.0% 30.8% 30.3% 11,924 * Post 17.1% 31.0% 30.8% 33,086 * Panel C. Two-Sided-State Full 41.4% -2.0% 19.3% 92,366 * Pre 45.9% 19.4% 19.2% 33,896 * Post 39.7% 22.9% 19.3% 63,478 * Note: C = M = National Stock Exchange, D Automated Display Facility, I International Stock Exchange, Chicago Stock Exchange, N = NYSE, P = ARCA/Pacific, T = NASDAQ, X = Philadelphia Stock Exchange, and W = CBOE. * Significant at the 0.10 level. Table XI. Exchange Quoting Behavior We evaluate quoting behavior at the NBBO Quote and at Benchmark Compliant Prices for the overall market during Benchmark Events and present the results here. We compute our gap measures on each side of the market separately. Max Ask Gap is the distance in cents between the NBBO Ask and the Benchmark Ask. Ask Gap is the difference in cents between a given Benchmark Compliant Price on the ask side and the NBBO Ask. The first row reports Benchmark Events with a Max Gap of $.01 (ask or bid). During this Benchmark Event, liquidity suppliers at an exchange can quote at the NBBO Quote, which has a Gap of zero, or at the Benchmark Ask (Bid), which has an Ask (Bid) Gap of one. The second row presents the results for Benchmark Events with a Max Gap of $.02 when liquidity suppliers can quote at the NBBO Quote (Gap = 0), at the Benchmark Ask (Bid) (Ask (Bid) Gap = 2), or at a Benchmark Compliant Price (Ask (Bid) Gap = 1). We continue in this fashion for each of the remaining rows. Next, we repeat the analysis on the bid side. Since the predictions for each side are the same, we combine the ask- and bid-side values and report and test the two sides combined. Thus, the results are for the three states in which a Benchmark Event occurred. Column 3 presents the number of exchanges quoting at Benchmark Compliant Prices and Column 4 presents Total Breadth, which is a count of the number of exchanges quoting at the NBBO Quote and at Benchmark Compliant Prices. On a time weighted basis, when the Max Gap is one, 2.26 exchanges quote at the NBBO Quote and 1.73 exchanges quote at the Benchmark Quote. The count of Benchmark Events that are initiated by an increase in the Max Gap is presented in Benchmark Events Column. %NBBO Breadth column in Panel A has a Pearson correlation coefficient with Max Gap of -0.360, indicating that they have a statistically significant downward trend at the 0.01 level. The same correlation for Panel B is -0.370. The results in Panel A are for the entire sample and in Panel B for Benchmark Events that exceed 50 milliseconds. Breadth Max Gap (in cents) NBBO Benchmark Total Panel A. Full Sample 1 2.26 1.73 3.99 2 1.91 2.08 3.99 3 1.68 2.19 3.87 4 1.56 2.16 3.71 5 1.45 2.10 3.54 6 1.36 2.02 3.38 7 1.28 1.95 3.22 8 1.19 1.84 3.03 9 1.11 1.76 2.88 10+ 1.20 1.85 3.05 Panel B. Benchmark Events Exceeding 50 Milliseconds 1 2.28 1.72 4.00 2 1.92 2.08 4.00 3 1.68 2.19 3.88 4 1.56 2.16 3.72 5 1.45 2.10 3.55 6 1.37 2.03 3.39 7 1.28 1.95 3.23 8 1.19 1.84 3.03 9 1.11 1.76 2.88 10+ 1.20 1.85 3.05 Max Gap %NBBO %Benchmark Benchmark (in cents) Breadth Breadth Events Panel A. Full Sample 1 0.556 0.444 5,353 2 0.478 0.522 2,419 3 0.435 0.565 1,147 4 0.421 0.579 672 5 0.412 0.588 401 6 0.410 0.590 268 7 0.406 0.594 179 8 0.407 0.593 128 9 0.403 0.597 91 10+ 0.404 0.596 348 Panel B. Benchmark Events Exceeding 50 Milliseconds 1 0.559 0.441 4,923 2 0.478 0.522 2,289 3 0.435 0.565 1,094 4 0.421 0.579 641 5 0.412 0.588 382 6 0.410 0.590 256 7 0.405 0.595 170 8 0.406 0.594 122 9 0.402 0.598 87 10+ 0.404 0.596 332
|Printer friendly Cite/link Email Feedback|
|Author:||McInish, Thomas H.; Upson, James|
|Article Type:||Statistical data|
|Date:||Sep 22, 2013|
|Previous Article:||Institutional investment horizons and the cost of equity capital.|
|Next Article:||The timing of opening trades and pricing errors.|