Regulation NMS and market quality.

We show that both the quoted and effective spreads increased, the quoted depth decreased, and the market quality index decreased after the implementation of Regulation National Market System (NMS) (Reg NMS). We also find an increase in the price impact of trades and the dispersion of the pricing error after Reg NMS. The order execution speed is slower, the order fill rate is lower, and the order cancellation rate is higher for most trades after Reg NMS. Hence, contrary to the Securities and Exchange Commission's belief Reg NMS has proven to be detrimental to most traders. NASDAQ provided faster and more reliable executions than the NYSE/AMEX, and NASDAQ gained market shares from the NYSE/AMEX and other trading venues after Reg NMS.

In this study we examine the impact of regulation national market system (NMS) (Reg NMS), arguably one of the most comprehensive and controversial regulatory changes in the US financial markets in the last 30 years, on various dimensions of market quality using data before and after its implementation. (1) In particular, we analyze the effects of the two new rules, the order protection rule (OPR) and the access rule, on execution cost, execution speed and probability, price impact, and the efficiency of price discovery. Although there have been debates on the likely effects of Reg NMS on market quality and investor welfare, empirical evidence on the issue is scanty. To our best knowledge, the present study is the first to provide such evidence.

The OPR requires trading centers to "establish, maintain, and enforce written policies and procedures reasonably designed to prevent the execution of trades at prices inferior to protected quotations displayed by other trading centers, subject to an applicable exception." (2) OPR differentiates markets into fast and slow. Electronic markets are considered fast markets, while manual, floor-based exchanges are considered slow markets. Fast markets are not allowed to trade-through better prices on other fast markets, but are allowed to trade-through better prices on slow markets. Today, more than thirty venues compete for US equity trades, including primary and regional exchanges, crossing networks, Electronic Communication Networks (ECNs), and dark pools. OPR is prompted, in large part, by the Securities and Exchange Commission's (SEC's) concern that the increased fragmentation of trading and quoting across venues may reduce liquidity. In particular, the SEC fears that brokers executing trades in one market may trade-through better quotes in other markets, reducing the incentive to post the best possible quotes. (3) If they get traded-through frequently, liquidity providers may be less willing to supply liquidity, reducing overall market liquidity.

The access rule (AR) requires fair and nondiscriminatory access to quotations displayed by self-regulatory organization (SRO) trading centers through private linkages. AR complements OPR as it helps protect the best displayed quotes against trade-through by allowing broker-dealers and trading centers to access those quotes easily and cheaply. AR also increases the accuracy of displayed quotations by establishing an upper bound on the cost (i.e., the access fee) of accessing such quotations. (4) In addition, AR requires each SRO to adopt, maintain, and enforce rules that will prevent their members from displaying quotations that lock or cross the protected quotations of other trading centers. AR allows trading centers to display automated quotations that lock or cross the manual quotations of other trading centers. Thus, the rule recognizes the disparity in speed of response between automated (fast) and manual (slow) quotations.

The SEC believes that the protection of public limit orders provided by OPR would help reward liquidity suppliers and encourage competition among traders, thus increasing market liquidity and reducing trading costs. The SEC also believes that strong intermarket price protection offers greater assurance that investors who submit market orders receive the best available prices. (5) However, others disagree. Blume (2002, 2007) and O'Hara (2004) hold that Reg NMS does not properly recognize the diversity and differential needs of traders. O'Hara (2004) suggests that OPR may lead to a deterioration of liquidity as some traders may bypass better quotes on the NYSE for speedier trades on an automated system.

Our study sheds some light on these debates by comparing various measures of liquidity and market quality between the pre- and post-NMS periods, after controlling for changes in stock attributes surrounding its implementation. To measure the net effect of Reg NMS on liquidity and market quality after controlling for the effect of the credit market crisis and other market-wide changes, we also employ a difference-in-difference approach using the control group of stocks that are similar to the test sample, but were not subject to Reg NMS at the same time. In addition, we analyze the changes in execution speed and probability and how these changes affected market shares of different venues.

We show that both the quoted and effective spreads increased, the quoted depth decreased, and the market quality index decreased after the implementation of Reg NMS. Spreads increased across all trade size categories and the largest trade size group exhibited the largest increase, indicating that Reg NMS had a greater adverse effect on larger traders. We also find an increase in the price impact of trades and the dispersion of the pricing error after Reg NMS. Our results show that the order execution speed is slower in the post-NMS period for most market orders and marketable limit orders. The order fill rate decreased across all order types and the order cancellation rate increased for both marketable and nonmarketable limit orders after the implementation of Reg NMS. The decrease in the fill rate and the concurrent increase in the order cancellation rate may have occurred as more high-frequency traders (HFTs) split up orders to hide intentions and cancelled orders to maneuver around.

Overall, our results indicate that the post-NMS period exhibits generally poorer market quality for both NYSE/AMEX and NASDAQ stocks in terms of larger trading costs, greater pricing errors, slower order execution speeds, and lower execution probability. Hence, contrary to the SEC's belief, Reg NMS has proven to be detrimental to most traders (especially institutional traders), as many scholars and market participants have predicted. Finally, we find that NASDAQ provided faster and more reliable executions than the NYSE/AMEX, and NASDAQ gained additional market shares from the NYSE/AMEX and other trading venues after the implementation of Reg NMS.

Recently, market observers (Hendelman and Rowley, 2010; Redler, 2010) suggest that the SEC should rethink and revise Reg NMS. For instance, Redler (2010) argues that OPR should be repealed as it led to an explosion of HFTs and a deterioration of market liquidity:

"The inability to pay through the inside market has removed any incentives to sit outside with a block of liquidity. It forces traders to pay the small players that create a tighter inside spread. Yet, the "real" spread in terms of sizable, reliable liquidity may be a few cents out on either side. The inability to simultaneously execute orders through the inside market allows HFTs to layer stocks and then rapidly cancel upon the first transaction. Relaxing the Order Protection Rule would allow traders to go after the real liquidity in the market and force greater reliability of orders that are placed outside the inside market. "

Our empirical results are consistent with these observations. Our results are also consistent with the prediction that OPR will reduce market liquidity because it reduces the role of NYSE specialists and floor brokers as both the liquidity providers of last resort (Hendelman and Rowley, 2010; Mehta, 2010) and information intermediaries (Benveniste, Marcus, and Wilhelm, 1992; O'Hara, 2004; Hendershott and Moulton, 2011) as more traders may bypass the NYSE for speedier trades on an automated system. Based on these results, we concur with those market observers who have suggested that the SEC may need to revisit and revise Reg NMS.

The remainder of the paper is organized as follows. Section I summarizes major elements of debates regarding the possible effects of Reg NMS on market quality and investor welfare. Section II describes data sources and variable measurements. Sections III-VI report our findings about the effects of Reg NMS on various market quality measures. Section VII shows how Reg NMS has affected market shares of different trading venues. Finally, Section VIII presents a brief summary and concluding remarks.

I. Regulation NMS and Market Quality

Appendix A summarizes key differences in trade-through provisions between the intermarket trading system (ITS) and the OPR and arguments for and against OPR. Appendix B summarizes key differences between the pre and post-AR and arguments for and against AR.

Much of the debate regarding Reg NMS centers on issues related to trade-through. Some advocated the elimination of any trade-through rule, while others continued to support trade-through protection. Statements from the adopting release itself illustrate the controversy about Reg NMS. The Commission's majority opinion is that"the rules [of Regulation NMS] are designed to assure that the equity markets will continue to serve the interests of investors, listed companies, and the public for years to come." The dissenting Commissioners' opinion is that "the majority's statutory interpretations and policy changes are arbitrary, unreasonable, and anticompetitive." (6) The 3-2 SEC vote indicates the divisiveness in the member opinion. The majority view of the SEC is that stronger protection of public limit orders (by prohibiting trade-throughs) would help reward liquidity suppliers and encourage competition among traders, thus increasing market liquidity.

Reg NMS was prompted, in large part, by the SEC's belief that market fragmentation reduces liquidity and that the new regulation would help create a more integrated market. Prior research offers useful insights on the effect of market fragmentation/integration on trader behavior and market quality. Madhavan (1995) predicts that some market participants (e.g., large traders) prefer to trade in a fragmented market where their trades are not disclosed. Hendershott and Jones (2005) find evidence that an increase in market fragmentation leads to slower price discovery and larger trading costs. Barclay, Hendershott, and Jones (2008) find that the consolidation of orders is important for producing efficient prices, especially during times of high liquidity demand. In contrast, Foucault and Menkveld (2008) and O'Hara and Ye (2011) present evidence that fragmentation of order flow actually enhances liquidity supply, reduces transactions costs, and increases execution speeds, suggesting that Reg NMS may not have its intended effect on market liquidity. (7)

A group of researchers suggest that Reg NMS would not improve execution quality as it does not recognize the diversity of traders. (8) In particular, they argue that trade-through prohibitions are detrimental to institutional traders since quoted depths at the inside market are usually much smaller than typical institutional trade sizes. Without the ability to opt out, trade-through prohibitions prevent institutional investors from accessing large amounts of liquidity at prices that are slightly worse than the inside quote, thereby reducing execution speed and increasing execution costs.

Blume (2002, 2007) and O'Hara (2004) raise a question regarding the merit of the implicit presumption of Reg NMS that price is the foremost standard of execution quality as it does not recognize the differential needs of traders. For instance, some investors may be willing to pay larger execution costs in return for faster executions or greater size. The price impact of trades is likely to be a much more important concern to institutional investors than to individual traders. Blume (2002, 2007) also argues that Reg NMS will negate innovative developments and deter competition between markets because it imposes the SEC's views regarding how the markets should be structured and how equities should be traded.

O'Hara (2004) suggests that OPR would increase automated trades and reduce market liquidity. (9) Prior to Reg NMS, trade-throughs were not allowed in all (fast and slow) markets. O'Hara (2004) underscores that under OPR, some traders would bypass the NYSE for speedier trades on an automated system even if the price available on the automated system is poorer than the specialist quote since OPR allows fast markets to trade-through slow markets. Consistent with this prediction, the specialist participation on the NYSE decreased significantly during the phase-in period of Reg NMS. (10) The author holds that insofar as the order flow diversion from the NYSE is large, OPR would have a detrimental effect on liquidity given the finding of prior research that the NYSE provides the lowest spread and price impact (a result that the NYSE would attribute to its specialist system). (11) Blume (2007) and O'Hara (2004) also suggest that OPR can adversely affect price discovery and market liquidity as it may increase internalized orders at the large dealer firms, removing these orders from a public market. (12)

II. Data Sources and the Measurement of Variables

Our study sample consists of NYSE, American Stock Exchange (AMEX), and NASDAQ securities. We exclude American Depository Receipts, mutual funds, warrants, firms in bankruptcy proceedings, firms delinquent in required filings with the SEC, and foreign companies from our study sample. We retrieve trade and quote data from the NYSE's Trade and Quote (TAQ) database and the number of shares outstanding from the Center for Research in Security Prices (CRSP) database. We construct national best bid and offer (NBBO) quotes for our study sample of NYSE, AMEX, and NASDAQ stocks during periods before and after the implementation of Reg NMS. We omit the following quotes and trades to minimize data error: 1) quotes if either the ask or bid price is nonpositive, 2) quotes if either the ask or bid size is nonpositive, 3) quotes if the bid-ask spread is greater than $5 or nonpositive, 4) before-the-open and after-the-close trades and quotes, 5) trades if the price or volume is nonpositive, 6) bid quote, [Bid.sub.t], if [([Bid.sub.t] - [Bid.sub.t-1])/[Bid.sub.t-1][ > 0.5, 7) ask quote, [Ask.sub.t], if [([Ask.sub.t] - [Ask.sub.t-1])/[Ask.sub.t-1]| > 0.5, and 8) trade price, [p.sub.t], if [absolute value of ([P.sub.t] -[p.sub.t-1])/[p.sub.t-l]] > 0.5.

We measure the quoted and effective spreads, market quality index, and price impact using the following formulas: (13)

Quoted dollar [spread.sub.i,t] = [Ask.sub.i,t] - [Bid.sub.i,t], (1)

Quoted percentage [spread.sub.i,t] = ([Ask.sub.i,t] - [Bid.sub.i,t])/[M.sub.i,t], (2)

Effective dollar [spread.sub.i,t] = 2[D.sub.i,t]([P.sub.i,t] - [M.sub.i,t]), (3)

Effective percentage [spread.sub.i,t] = [2D.sub.i,t]([P.sub.i,t] - [M.sub.i,t] ) / [M.sub.i,t], (4)

Market quality [index.sub.i,t] = 1/2(Bid [size.sub.i,t] + [Asksize.sub.i,t])/[([Ask.sub.i,t] - [Bid.sub.i,t])/[M.sub.i,t],], (14) (5)

Price [impact.sub.i,t] = [D.sub.i,t]([M.sub.i,t+5] - [M.sub.i,t]), (6)

where [Ask.sub.i,t] is the national best ask price of stock i at time t, [Bid.sub.i,t] is the national best bid price of stock i at time t, [M.sub.i,t] is the quote midpoint ([[Ask.sub.i, t] + [Bid.sub.i, t]]/2) of stock i at time t, [M.sub.i, t+5] is the quote midpoint at time t + 5 minutes of stock i, [P.sub.i,t] is the transaction price of stock i at time t, Bid [size.sub.i,t] is the quoted size at the bid of stock i at time t, Ask [size.sub.i,t] is the quoted size at the ask of stock i at time t, and [D.sub.i,t] is an indicator variable equal to + 1 for customer buy orders and - 1 for customer sell orders. We estimate [D.sub.i.t] using the algorithm in Lee and Ready (1991) and Bessembinder (2003) with no allowance for trade reporting lag. For each stock, we calculate the time-weighted mean quoted spread and trade-weighted mean effective spread. We measure the dollar depth of each stock by the sum of dollar bid size and dollar ask size. We measure return volatility by the standard deviation of daily quote midpoint returns. We also calculate the daily number of trades and the average dollar trade size for each stock.

Both AR and OPR were first implemented on July 9, 2007 for a pilot sample of 250 NMS stocks (i.e., 100 NYSE stocks, 100 NASDAQ stocks, and 50 AMEX stocks). (15) The primary listing market, in consultation with the SEC, selected the pilot samples that are reasonably representative. The NYSE/AMEX and NASDAQ pilot samples comprise stocks in various industries including mining, manufacturing, transportation and public utilities, wholesale trade, retail trade, finance, insurance, real estate, and services. In June 2007, the average daily trading volume for NYSE/AMEX pilot stocks ranged from $9,000 to $1.6 billion. The market value of equity for the same month ranged from $15 million to $390 billion. For NASDAQ pilot stocks, the daily trading volume (market value of equity) ranged from $240,000 ($40 million) to $86 million ($9 billion).

The main purpose of the pilot phase was to allow market participants to verify the functionality of their policies, procedures, and systems that were necessary to comply with the Reg NMS requirements. For the remaining NMS stocks, the full industry compliance of the rules began on August 20, 2007 and was completed on October 8, 2007. Because these two (pilot and main) groups of stocks have different compliance dates, we perform empirical analyses using data for each group separately. After omitting stocks with incomplete data, our final study sample consists of 98 NYSE, 48 AMEX, and 96 NASDAQ stocks for the pilot group and 2,343 NYSE, 837 AMEX, and 2,757 NASDAQ stocks for the main group.

The impending migration of orders to the best automated quote foreshadowed by the Reg NMS proposals had exerted an impact on markets well before the rule implementation. From August 2002 to June 2007 throughout the five-year dialogue about Reg NMS, NYSE automation grew and specialist participation dropped from 15.1% to 3.1%. 16 These effects bias the results against finding any effect of Reg NMS on market quality.

[FIGURE 1 OMITTED]

III. Empirical Results

In this section, we analyze the effect of Reg NMS on market quality using data prior to and after its implementation. (17)

A. Results for the Pilot Group

To examine the effects of Reg NMS on the spread, depth, market quality index, price impact, and other attributes of pilot stocks, we compare the mean value of each variable between the preand post-NMS periods. Although the quoted depth in the TAQ database does not truly reflect the total available liquidity for NASDAQ stocks, we include the quoted depth and market quality index in our analysis for completeness. Since the implementation date for pilot stocks is July 9, 2007, we use 30 trading days before July 9, 2007 (i.e., May 24, 2007 to July 6, 2007) as the pre-NMS period and 30 trading days from July 9, 2007 (i.e., July 9, 2007 to August 17, 2007) as the post-NMS period (see Figure 1). (18)

Table I reports the results for the 242 pilot stocks. The results indicate that the average quoted percentage spread increased by 0.0004 (a 21.05% increase) from 0.0019 in the pre-NMS period to 0.0023 in the post-NMS period. Similarly, the average effective percentage spread increased by 0.0004 (a 28.57% increase) from 0.0014 to 0.0018. We find qualitatively similar results for the quoted and effective dollar spreads. The dollar depth declined by $26,017 (a 31.88% reduction) from $81,618 to $55,601 and the market quality index declined by 0.8663 (a 36.94% reduction) from 2.3454 to 1.4791. Return volatility increased from 0.001 to 0.0013 (a 30% increase) and the price impact of trades increased by 0.0048 (a 36.36% increase) from 0.0132 to 0.0180. The higher return volatility and the larger price impact in the post-NMS period may be due, at least in part, to reduced market depths. The number of trades increased by 2,921 (a 37.51% increase) from 7,786 to 10,707, while the average trade size decreased by $1,566 (a 15.3% decrease) from $10,236 to $8,670.

B. Results for the Main Group

The phase-in date of Reg NMS for all other (nonpilot) stocks is August 20, 2007. We use 30 trading days before August 20, 2007 (i.e., July 9, 2007 to August 17, 2007) as the pre-NMS period and 30 trading days from August 20, 2007 (i.e., August 20, 2007 to October 1, 2007) as the post-NMS period. (19) Figure 1 illustrates the timeline of Reg NMS for the main implementation group. Table I reports market quality measures for the main implementation group of 5,937 NYSE, AMEX, and NASDAQ stocks during the pre-NMS period and the post-NMS period, respectively.

The results for the main implementation sample are qualitatively similar to those for the pilot sample. For example, the average quoted percentage spread increased by 0.0008 (18.6%) from 0.0043 in the pre-NMS period to 0.0051 in the post-NMS period and the average effective percentage spread increased by 0.0005 (18.52%) from 0.0027 to 0.0032. The dollar trading volume on the NYSE was $21,216 billion in 2007. Assuming that traders pay half the effective spread as the execution cost for each share traded, an increase of 0.0005 (=0.0032 - 0.0027) in the effective percentage spread translates into an increase $5.3 billion ([=$21,216 billion x (0.0032/2- 0.0027/2]) in the annual execution cost on the NYSE alone.

Comparison of the results between the pilot and main groups indicates that the effects of Reg NMS on our market quality measures for the main group are smaller than the corresponding figures for the pilot group. Although the phase-in date of Reg NMS for nonpilot stocks is August 20, 2007, the effect of Reg NMS on some of these stocks may have started to show up around (or before) the implementation date for the pilot stocks if market participants built their systems well before the implementation date and began to comply with Reg NMS for all stocks on the same date. (20) They may also have found it cumbersome to differentiate stocks according to the rule compliance date. These factors may explain, at least in part, why the effects of Reg NMS on our market quality measures differ between the two groups of stocks.

We also compare spreads, dollar depths, market quality index, return volatility, and price impact between the pre- and post-NMS periods within each of the four trade size categories (i.e., fewer than 500 shares, 500-1,999 shares, 2,000-5,000 shares, and larger than 5,000 shares). The results are qualitatively similar to those reported in Table I. For example, we find that spreads increased after Reg NMS across all four categories and the largest trade size group exhibited the largest increase in spreads, indicating that Reg NMS had a greater adverse effect on larger traders. This result is consistent with the view of the opponents of OPR that trade-through prohibitions are more detrimental to institutional traders as quoted depths at the inside market are usually smaller than typical institutional trade sizes (these results are available from the authors upon request).

C. Are Our Results Driven by Concurrent Changes in the Market Environment?

Our empirical results in the previous section show that for NYSE/AMEX/NASDAQ securities, the quoted and effective spreads increased, the quoted dollar depth decreased, and the market quality index decreased significantly after Reg NMS. It is possible that these changes may be attributable to changes in the market environment surrounding the implementation of Reg NMS. For instance, as we showed above, both trading frequency and return volatility increased significantly after Reg NMS. (21) To examine whether the changes in the spread, depth, and market quality index are indeed due to Reg NMS after controlling for changes in the market environment, we estimate the following regression model:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (7)

where superscripts "post" and "pre" denote the post- and pre-NMS periods, respectively; [VAR.sub.i] denotes the quoted spread, effective spread, dollar depth, or market quality index of stock i; [X.sub.k] (k = 1-4) represents one of the four stock attributes (i.e., share price, dollar trading volume, return volatility, and return); [beta]s are the regression coefficients; and [[epsilon].sub.i] is the error term. In regression model (7), the intercept [[beta].sub.0] measures the difference in variables between the pre- and post-NMS periods after controlling for changes in the four stock attributes.

We report the regression results for the pilot sample ofNYSE, AMEX, and NASDAQ stocks in Panel A of Table II. Likewise, we report the regression results for the main implementation sample of NYSE, AMEX, and NASDAQ stocks in Panel B of Table II. For both the pilot and main groups, we find that the estimated intercept is similar to the difference in each market quality measure between the pre- and postperiod, indicating that the results in the previous two sections are not driven by changes in the stock attributes.

D. Discussion of Results

Overall, our results indicate that liquidity providers on the NYSE, AMEX, and NASDAQ post wider spreads and smaller depths after the implementation of Reg NMS. Hence, contrary to the SEC's expectation, OPR neither reduces transaction costs nor improves market liquidity by encouraging public limit orders. (22) Rather, our results support the view of opponents of the rule that Reg NMS would actually reduce market liquidity. Although it is difficult to isolate the exact cause of these results, we provide some possible explanations below.

Benveniste et al. (1992) find that a specialist who differentiates between informed and uninformed traders through his long-term relationship with brokers can achieve equilibrium that Pareto-dominate a pooling equilibrium in which he does not differentiate between the two types of traders. Their study provides an important insight that human intermediation is valuable in securities markets. Battalio, Ellul, and Jennings (2007) show that liquidity costs increase around a stock's relocation on the NYSE trading floor, and conclude that reputation plays an important role in the NYSE's liquidity provision process. Hendershott and Moulton (2011) suggest that the decrease in floor trading brought on by the hybrid market reduces human intermediation, thus increasing adverse selection costs. The lower liquidity after Reg NMS may be attributed, at least in part, to the reduced role of the specialist (and floor brokers) in handling information asymmetry problems as more traders bypass superior specialist quotes on the NYSE for speedier trades on an automated system.

Many market observers believe that Reg NMS caused the explosion of liFTs who take advantage of other traders' intention to buy or sell. (23) HFTs use computer programs to detect the footprints of larger players and trade off of the order flow for small gains. HFTs make money by trading alongside large traders rather than by taking the opposite side of the trade. HFTs also try to avoid large losses that could result from wild movements in the stock market. (24) In response, large buy-side firms have invested heavily in algorithm equipment to protect their order flow from HFTs. These firms have moved to dark pools that allow them to trade large blocks with each other without displaying the liquidity on public order books. This may be another reason for a deterioration of liquidity after the implementation of Reg NMS.

Although the growth in high frequency trading has been argued to have filled any liquidity void, the role of liFTs as liquidity providers is fundamentally different from that of the specialist. Throughout the history of the US equity markets, NYSE specialists were required to provide fair and orderly markets, especially during times of high volatility and low liquidity. NASDAQ dealers performed similar functions over phones and computers. In contrast, HFTs do not have comparable affirmative responsibilities to create markets and provide needed liquidity. The lower liquidity after Reg NMS may also be explained by the reduced role of NYSE specialists and floor brokers as the liquidity providers of last resort (Hendelman and Rowley, 2010; Mehta, 2010).

It is important to note that the NMS rule compliance required significant investments by each trading center in order management systems as well as order routing, market data repository, and compliance systems. For instance, trading centers needed to develop and implement written policies and procedures on how they supervise and survey trade-through and exception compliance. Trading centers also needed to programmatically code the trade-through rules within their order management and execution systems. In addition, trading centers needed to create the necessary private linkages to ensure that all protected quotes can be accessed, develop surveillance systems to monitor compliance by traders with the trade-through rules, and ensure the integrity of automated systems. Even after these initial investments, trading centers are likely to incur additional recurring costs to continuously satisfy the rule compliance requirement. To the extent trading centers need to recoup their initial investment and other recurring costs, liquidity providers who are affiliated with these trading centers may have to quote larger spreads than they did prior to the implementation of Reg NMS as a result of the larger order processing component of the spread.

Another possible explanation for the increased spread and reduced depth may be the newly imposed upper limit (i.e., $0.003) on access fees by AR. Access fees are likely to be greater when markets pay larger rebates to liquidity providers. For example, a number of ECN trading centers charge access fees to incoming orders that execute against their displayed quotations and pass a substantial portion of the access fee on to limit order customers as rebates for supplying the liquidity. To the extent that AR decreased the access fee and, consequently, reduced the rebate to liquidity providers, they might have increased spreads and reduced depths to recoup the reduced revenues from the rebate. Hence, the increase in spreads and the decrease in depths may be attributed, at least in part, to the newly imposed cap on access fees.

IV. Are the Results Driven by Other Concurrent Events?

US credit markets experienced a significant deterioration in the prices of mortgage-related products in 2007. Major credit rating agencies (e.g., Moody's, Standard & Poor's, and Fitch) downgraded a number of mortgage tranches in June and July of 2007 leading to a significant increase in risk premiums in the bond market. (25) Furthermore, Reg NMS became effective in the middle of the so-called "quant meltdown" (Khandani and Lo, 2007). To the extent that liquidity depends upon the financial condition of liquidity suppliers (Comerton-Forde et al., 2010), this event could have first order effects on market liquidity. (26) Other market-wide confounding events include the repeal of the uptick rule on July 6, 2007. (27)

To measure the net effect of Reg NMS on liquidity after controlling for the effect of the credit market crisis and other market-wide events, we employ a difference-in-difference approach using the control group of stocks. For the pilot sample, we use the matching sample of stocks from the main implementation group as the control sample (see Figure 1). Note that although both the pilot and main implementation groups of stocks are subject to the effect of the credit market crisis, only those stocks in the pilot group are subject to Reg NMS during the pilot implementation period (i.e., July 9, 2007-August 17, 2007). Consequently, we analyze the net effect of Reg NMS by testing whether the difference in liquidity measures between the pilot and control samples is greater during the post-NMS period. (28)

Stocks in the pilot sample and in the main implementation group could be different in dimensions that are related to liquidity measures. Following Huang and Stoll (1996) and Bessembinder and Kaufman (1997), we minimize such differences for a meaningful comparison of liquidity between these groups of stocks. Specifically, we obtain the control sample of stocks that are similar to the pilot sample in terms of price, return, trading volume, and return volatility within the same industry. For this, we first calculate the composite match score (CMS) of each pilot stock against each and every stock with the same two-digit standard industrial classification (SIC) code in the main implementation group: CMS = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.], where [X.sub.k] represents one of the four stock attributes and [SIGMA] denotes the summation over k = 1-4. Then, for each pilot stock, we select the stock in the main group with the lowest score. Once we match a main stock with a pilot stock, we no longer consider that particular main stock for subsequent matches. This procedure results in 146 matching pairs of pilot and main NYSE/AMEX stocks and 96 pairs of pilot and main NASDAQ stocks.

Table III reports the attributes of 242 (146 + 96) matching pairs of NYSE/AMEX and NASDAQ stocks. The results indicate that the average price, trading volume, return, and return volatility of the pilot sample are $33.08, $124.08 million, -0.0013, and 0.0010, respectively. The corresponding figures for the control group are $32.80, $122.61 million, -0.0014, and 0.0011. The average paired differences in stock attributes between the pilot and control stocks are $0.28 for price, $1.47 million for trading volume, 0.0001 for return, and -0.0001 for return volatility. Our t-test results indicate that none of these differences are statistically significant, indicating that our test and control samples of stocks are very similar in dimensions that are related to liquidity measures.

We estimate the following regression model using the pilot and control (matching) sample of stocks:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (8)

where [VAR.sub.i,t] is the quoted dollar spread, quoted percentage spread, effective dollar spread, effective percentage spread, dollar depth, or market quality index of stock i on day t, [D.sub.t.sup.NMS] is an indicator variable equal to one for post-NMS days (i.e., July 9, 2007 to August 17, 2007) and zero for pre-NMS days (i.e., May 24, 2007 to July 6, 2007), [X.sub.i,t,k] (k = 1-4) represents one of the four stock attributes of stock i on day t, and [SIGMA] denotes the summation over k = 1-4. The superscripts "pilot" and "control" represent the pilot and control stocks, respectively. Although our test (pilot) and control samples of stocks are similar to each other, small differences still exist between the two samples. We include ([X.sup.pilot.sub.i,t,k] - [X.sup.control.sub.i,t,k] in the regression model to control for such differences. Also included in the model are a matched pair fixed effect ([[lambda].sub.i]) and dummy variables for each trading day ([[theta].sub.t]). We include the matched pair fixed effect to control for any differences between two stocks in a pair that are present during the pre-NMS period. We include the time dummy variables to control for any market-wide changes in the dependent variable.

To assess the sensitivity of our results, we also employ the following three methods (see Petersen (2009) for a detailed description of these methods): (29)1) drop [[theta].sub.t] from regression model (8) and use standard errors clustered by firm; 2) drop [[lambda].sub.i] from regression model (8) and use standard errors clustered by time; and 3) drop both [[theta].sub.t] and [[lambda].sub.i] from regression model (8) and use standard errors clustered by firm and time. (30) The results from these alternative models are qualitatively similar to those from regression model (8). Thus, for brevity, we report only the results of regression model (8). (31)

We present the regression results in Panel A of Table IV. The results indicate that the estimated coefficients on [D.sub.t.sup.NMs] are positive and significant in the regression models for the quoted and effective spreads, and negative and significant in the regression models for the dollar depth and market quality index, respectively. These results indicate that the observed deterioration in liquidity after the pilot implementation of Reg NMS cannot be entirely attributed to the crisis in the credit market and/or other confounding events. Note that the estimated coefficients on [D.sub.t.sup.NMs] are all smaller in absolute value than the estimated regression intercepts in Panel A of Table II. This result suggests that at least part of the changes in liquidity and market quality measures (captured by these regression intercepts) after the implementation of Reg NMS could be attributed to the credit market crisis and/or other market-wide events.

We follow a similar method to measure the net effect of Reg NMS on the liquidity of the main implementation group of stocks. We use the pilot sample of stocks as the control group as these stocks have already been subjected to Reg NMS on July 9, 2007. Thus, the implementation of Reg NMS for the main group that began on August 20, 2007 should not have a material effect on the liquidity of the pilot sample during the post-NMS period (i.e., August 20, 2007 to October 1, 2007) (see Figure 1). Consequently, we analyze the net effect of Reg NMS by testing whether the difference in liquidity measures between the main and control samples is greater during the post-NMS period.

We then estimate the following regression model using the main and control sample of stocks:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (9)

where [D.sub.t.sup.NMS] is an indicator variable equal to one for post-NMS days (i.e., August 20, 2007 to October 1, 2007) and zero for pre-NMS days (i.e., July 9, 2007 to August 17, 2007), the superscripts "main" and "control" represent the main and control stocks, respectively, and all other variables are the same as previously defined in regression model (8). The regression results (see Panel B of Table IV) are qualitatively similar to those from the pilot sample. The estimated coefficients on DtNMS are positive and significant in the quoted and effective spread regression models, and negative and significant in the dollar depth and market quality index regression models.

V. The Effect of Reg NMS on Pricing Error

Impaired price discovery could cause market prices to deviate from fundamental values and create excessive short-term volatility. One of the intended purposes of Reg NMS is to raise the information efficiency of asset price by integrating all equity trades into a common computerized trading system, rewarding market participants who are willing to provide liquidity, and allowing fair and nondiscriminatory access to quotations. For instance, the SEC believes that OPR will promote market efficiency by more effectively integrating trading centers into a common trading system.

Hasbrouck (1993) decomposes security transaction prices into random-walk and stationary components and identifies the random-walk component as the efficient price. The residual stationary component, the difference between the efficient price and the actual transaction price (i.e., pricing error), is then interpreted as the implicit transaction cost. Hasbrouck (1993) suggests the dispersion of the pricing error (which measures how closely actual transaction prices track random walk) as a reasonable measure of market quality. (32)

To calculate the pricing error variance, we first estimate the following models:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (11)

where RET is the return calculated from the transaction prices, TRD is the signed trade, t indicates the time subscript, and [mu]'s are the disturbance terms. The vector moving average (VMA) representation expressed in terms of disturbances is:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (12)

We calculate the pricing error variance using:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (13)

where [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] and COV([mu]) is the disturbance covariance matrix. We measure market quality by [[sigma].sub.s].

To measure the net effect of Reg NMS on pricing error after controlling for any time trend in each venue's pricing error, we employ a difference-in-difference approach using the control sample of stocks, similar to the approach in Section IV. Table V presents the mean pricing error for the test group and the control sample. Panel A shows the results for the pilot group, while Panel B reports the results for the main implementation group. As in other tables, we show whether the difference in pricing error between the pre- and post-NMS period is statistically significant. More importantly, we also report whether the difference in pricing error between the pre- and post-NMS period for the test group is significantly greater than the corresponding figure for the control sample.

Panel A shows that for the pilot sample, the pricing error of NYSE/AMEX-listed stocks increased by 0.0007 from 0.0056 to 0.0063, and the magnitude of the increase is significantly larger than the corresponding value (0.0002) for the control sample. We find qualitatively similar results for the pilot group and control sample of NASDAQ stocks. Panel B shows that for the main implementation group, the pricing error of NYSE/AMEX stocks increased by 0.0037, which is significantly larger than the corresponding figure (0.0002) for the control sample. We find similar results for the main implementation group and control sample of NASDAQ stocks. These results suggest that contrary to the SEC's expectation, Reg NMS led to a decrease in pricing efficiency. The lower pricing efficiency in the post-NMS period may be attributed, in part, to increased opportunistic trading by HFTs.

VI. Reg NMS and SEC Rule 605 Execution Quality Measures

The SEC adopted Rule 605 on November 15, 2000 to improve public disclosure of execution quality. Under Rule 605, market centers are required to make the monthly disclosure of execution quality for each stock. In this section, we analyze the effect of Reg NMS on Rule 605 execution quality measures. (33) We collect the Rule 605 data from the website of transaction auditing group. The pre- and post-NMS periods for the 605 execution quality data are May/June 2007 and September/October 2007, respectively.

We conduct the pre- and post-Reg NMS comparison of the effective spread and the price impact of trades for both market orders and marketable limit orders using Rule 605 data to assess whether the results we obtain from the TAQ data are robust. We compare other execution quality measures to shed additional light on the effect of Reg NMS. Many scholars believe that OPR reduces execution speed as it would slow down large institutional trades. (34) Alternatively, OPR may increase execution speed if it increases automated trades. We compare the execution speed of market orders and marketable limit orders between the pre- and post-Reg NMS periods to provide empirical evidence on these different views. We also examine whether Reg NMS led to an increase in the order cancellation rate and a decrease in the order fill rate as many market observers have suggested.

A. Comparison of Execution Quality between the Pre-and Post-NMS Periods by Order Types

Panel A of Table VI compares the execution quality of stocks in the main implementation group between the pre- and post-NMS periods. (35) We report the following execution quality measures: 1) the effective dollar spread for executions of covered orders, 2) the price impact for executions of covered orders (as measured by the difference between the effective dollar spread and the realized dollar spread), 3) the proportion of shares that are executed at the quote with price improvement and outside the quote, respectively, 4) the share-weighted average duration of time in seconds from the time of order receipts to the time of order execution for shares executed at the quote with price improvement and outside the quote, respectively, and 5) the proportion of shares that are executed at the receiving market center (fill rate) executed at other venues (away rate) and canceled prior to execution (canceled). We calculate the mean value of each variable for each stock during the periods prior to and after the implementation of the Reg NMS. Because transaction auditing group does not provide the Rule 605 execution quality data on market orders for NASDAQ stocks, we use only NYSE stocks to measure the execution quality of market orders. We use data for NYSE, AMEX, and NASDAQ stocks to measure the execution quality of marketable and nonmarketable limit orders.

The results indicate that the mean effective dollar spread in the post-NMS period is significantly larger than the corresponding figure in the pre-NMS period for both market orders and marketable limit orders. Similarly, the mean price impact of trades in the post-NMS period is significantly greater than the corresponding figure in the pre-NMS period for both types of orders. These results are consistent with the results from the TAQ database that we reported in Table I, indicating the robustness of our findings. (36)

The results show that the proportion of shares that received price improvement increased after Reg NMS for both market orders and marketable limit orders. The order execution speed is slower after Reg NMS for most market orders and marketable limit orders. For marketable limit orders, the average order execution time is about 2.4 and 6.1 seconds longer for orders that are executed at the quote and outside the quote, respectively, in the post-NMS period. We find qualitatively similar results for market orders that are executed at the quote and outside the quote, accounting for more than 95% of executed orders. These results support the view that OPR reduces the execution speed of large institutional orders.

We find that the fill rate is significantly lower in the post-NMS period, regardless of order type. For example, in the pre-NMS period, the receiving market centers executed 79.61% of all incoming market orders (in number of shares). The corresponding figure declined to 73.76% in the post-NMS period. The decrease in the fill rate of 5.85% is statistically significant. We find qualitatively similar results for both marketable and nonmarketable limit orders. Not surprisingly, the fill rate for nonmarketable limit orders is much smaller than the corresponding figure for marketable limit orders. The results indicate that the proportion of marketable and nonmarketable limit orders (0.3527 and 0.7731) that are canceled before execution in the post-NMS period are significantly larger than the corresponding figures (0.3416 and 0.7221) in the pre-NMS period. As expected, the proportion of shares that are canceled before execution is largest for nonmarketable limit orders (0.7221; 0.7731), followed by marketable limit orders (0.3416; 0.3527), and then by market orders (0.0036; 0.0025) during both the pre- and post-NMS periods. The decrease in fill rates and the concurrent increase in order cancelation rates may have occurred as more HFTs split up orders to hide intentions and canceled orders to maneuver around after the implementation of Reg NMS.

B. Execution Quality of NYSE/AMEX Stocks on the NYSE/AMEX and NASDAQ

To shed some light on whether the NYSE/AMEX provide better or poorer executions than NASDAQ, we compare the execution quality of NYSE/AMEX stocks between those trades that occurred on the NYSE/AMEX and those trades that occurred on NASDAQ using both the preand post-NMS data. (37) The results (Panel B of Table VI) indicate that the NYSE provided better executions than NASDAQ in terms of smaller effective spreads and price impact in the pre-NMS period, but poor executions than NASDAQ in terms of larger effective spreads and price impact in the post-NMS period. In both the pre- and post-NMS periods, NASDAQ provided better execution than the NYSE in terms of a higher proportion of trades that received price improvement, a lower proportion of trades that are executed outside the quote, and faster execution speed. Recall that one of the reasons that retail brokers were against Reg NMS was that routing orders to NASDAQ has given them faster execution than the NYSE. Our results show that NASDAQ continues to provide faster executions for marketable limit orders than the NYSE/ AMEX. (38)

The results also show that for both marketable and nonmarketable limit orders, the average fill rate on NASDAQ is significantly lower than the corresponding figure on the NYSE/AMEX in the pre-NMS period. However, the average fill rate on NASDAQ is significantly higher than the corresponding figure on the NYSE/AMEX in the post-NMS period. NASDAQ exhibits lower order cancellation rates for marketable limit orders than the NYSE in both the pre- and post-NMS periods. The order cancellation rate is also lower on NASDAQ for nonmarketable limit orders in the post-NMS period.

The last column of Panel B in Table VI shows that NASDAQ exhibits better executions in the post-NMS period than in the pre-NMS period in most market quality metrics with smaller effective spreads, smaller price impact, a greater proportion of price-improved trades, a lower proportion of trades that are executed outside the quote, faster execution speed, and higher fill rates. However, the results also indicate that order cancellation rates on NASDAQ increased for both marketable and nonmarketable orders after the implementation of Reg NMS.

As a whole, our results show that NASDAQ provided better executions in most market quality dimensions than the NYSE in the post-NMS period. (39) NASDAQ also provided better executions in the post-NMS period than in the pre-NMS period in these market quality metrics, except for higher order cancellation rates. The higher order cancellation rate in the post-NMS period may be attributable to the higher high frequency trading.

VII. The Impact of Reg NMS on Market Share

In this section, we examine the effect of Reg NMS on the market shares of the NYSE/AMEX, NASDAQ, and other trading venues (e.g., Boston, Philadelphia, Chicago, ARCA, NASD ADF and TRE ISE, NSX, and CBOE). (As of March 5, 2007, NASDAQ trades of NYSE-, AMEX-, and ARCA-listed stocks are reported in the TAQ database with an exchange code of T.) Reg NMS focuses on price as the foremost determinant of execution quality. As a result, traders are more likely to send orders to the market that offers a fast and sure execution in the post-NMS period than in the pre-NMS period. As more marketable orders are sent to the faster market, more limit orders would follow resulting in a greater market share. As noted above, NASDAQ provided faster executions and higher fill rates than the NYSE/AMEX for marketable limit orders during the post-NMS period and its post-NMS execution speed and fill rates are better than their pre-NMS values. Based on these considerations, we conjecture that Reg NMS leads to an increase in the market share of NASDAQ and a decrease in the market share of the NYSE/AMEX. Similarly, we expect an increase in the market share of ATSs (e.g., ARCA).

To determine the effect of Reg NMS on market share, we calculate the proportion of each stock's trades that are executed on the NYSE/AMEX, NASDAQ, and all other trading venues (e.g., ARCA) in the periods before and after the implementation of Reg NMS. To measure the net effect of Reg NMS on market share after controlling for any time trend in each venue's market share, we employ a difference-in-difference approach using the control sample of stocks as we did in Section IV.

Table VII presents the mean proportion of NYSE/AMEX-listed stocks' trades that are executed on the NYSE/AMEX, NASDAQ, and other trading venues for the test group and control sample. Moreover, it reports the mean proportion of NASDAQ-listed stocks' trades that are executed on NASDAQ and other trading venues for the test group and control sample. Panel A provides the results for the pilot group, while Panel B displays the results for the main implementation group.

Panel A indicates that for the pilot sample, the proportion of NYSE/AMEX-listed stocks' trades that are executed on the NYSE/AMEX decreased by 0.0471 from 0.4592 in the pre-NMS period to 0.4121 in the post-NMS period, and the magnitude of the decline (-0.0471) is significantly larger than the corresponding value (-0.0104) for the control sample. In contrast, the proportion of NYSE/AMEX-listed stocks' trades that are executed on NASDAQ increased by 0.0173 from 0.3945 to 0.4118, and the magnitude of the increase (0.0173) is significantly larger than the corresponding value (0.0058) for the control sample. Similarly, the proportion of NYSE/AMEX-listed stocks' trades that are executed on other trading venues increased by 0.0298 from 0.1464 to 0.1762, and the magnitude of the increase (0.0298) is significantly larger than the corresponding figure (0.0046) for the control sample. Panel A also shows that the proportion of NASDAQ-listed stocks' trades that are executed on NASDAQ (other venues) increased (decreased) by 0.0112 from 0.8224 (0.1765) in the pre-NMS period to 0.8336 (0.1653) in the post-NMS period, and the magnitude of the increase (decrease) is significantly larger than the corresponding figure for the control sample.

These results indicate that NASDAQ and other trading venues have gained market shares on the NYSE/AMEX-listed pilot stocks at the expense of the NYSE/AMEX after the implementation of Reg NMS. These results are consistent with our conjecture that traders are more likely to send orders to NASDAQ (which offers a faster and higher probability of execution than the NYSE/AMEX) in the post-NMS period than in the pre-NMS period. For NASDAQ pilot securities, NASDAQ increased its market share while other trading venues lost their market shares to NASDAQ after the implementation of Reg NMS. This result may also be the result of NASDAQ offering a faster and higher probability of execution than other trading venues.

Panel B of Table VII shows that for the main implementation group, the proportion of NYSE/AMEX-listed stocks' trades that are executed on the NYSE/AMEX decreased from 0.4466 in the pre-NMS period to 0.4105 in the post-NMS period, and the magnitude of the decrease (-0.0361) is significantly larger than the corresponding value (-0.0133) for the control sample. In contrast, the proportion of NYSE/AMEX-listed stocks' trades that are executed on NASDAQ increased from 0.3982 to 0.4195, and the magnitude of the increase (0.0213) is significantly larger than the corresponding figure (0.009) for the control sample. The proportion of NYSE/AMEX-listed stocks' trades that are executed on other venues also increased from 0.1550 to 0.1698, and the magnitude of the increase (0.0148) is significantly larger than the corresponding figure (0.0043) for the control sample. For NASDAQ-listed stocks, the proportion of trades that are executed on NASDAQ increased from 0.8297 in the pre-NMS period to 0.8494 in the post-NMS period, and the magnitude of the increase (0.0197) is significantly larger than the corresponding figure (0.0069) for the control sample, whereas the proportion of trades that are executed on other venues decreased from 0.1709 to 0.1512, and the magnitude of the decrease (-0.0197) is significantly larger than the corresponding figure (-0.0069) for the control sample. These results are qualitatively similar to those from the pilot sample.

Our empirical results in Table VII indicate that Reg NMS adversely affected the NYSE and AMEX as they lost their trading volume to NASDAQ and other trading venues after the implementation of Reg NMS. The results also indicate that the NASDAQ stock market gained additional market shares from other trading venues. Supporters of OPR (i.e., opponents of the old trade-through rule) suggested that the old rule confers monopoly status on the NYSE trading floor by requiring that orders in NYSE-listed stocks be exposed to the floor for 30 seconds to see if a better price can be found before they can be executed in other markets, including ATSs. The old trade-through rule negates the advantages of ATS trading (e.g., the speed, anonymity, and certainty of execution). Our results are in line with the view of the supporters of OPR that the 30-second exposure rule is anticompetitive as it grants an unfair advantage to the NYSE trading floor.

VIII. Summary and Concluding Remarks

Reg NMS invoked heated debates among regulators, scholars, and various market participants. Some argue that the new regulation would increase competition and reduce trading costs, while others express concerns regarding its possible negative consequences. In the present study, we shed some light on the debate by comparing various measures of market quality prior to and after the implementation of Reg NMS for a large number of NYSE-, AMEX-, and NASDAQ-listed stocks. We perform separate analysis for the pilot sample of stocks and the main implementation group of stocks.

Our results show that the effects of Reg NMS on market quality are qualitatively identical between the pilot and main implementation groups of stocks. We find that both the quoted and effective spreads increased and the quoted dollar depth decreased significantly after the implementation of Reg NMS. We find evidence that Reg NMS had a greater adverse effect on larger traders. We also find a higher price impact of trades and greater transitory price movements (i.e., pricing error) in the post Reg NMS period. Overall, these results suggest that Reg NMS resulted in greater trading costs, smaller market depths, and lower market efficiency. We also find evidence of slower execution speed, lower order fill rates, and higher order cancellation rates for the majority trades after the implementation of Reg NMS. Thus, Reg NMS appears to have adversely affected liquidity and execution quality as many scholars and market participants have predicted.

We find that NASDAQ exhibits better execution quality in terms of both faster execution speeds and higher execution probability than the NYSE/AMEX in the post-NMS period. NASDAQ also provided better executions in the post-NMS period than in the pre-NMS period in these market quality metrics. NASDAQ gained additional market shares from the NYSE/AMEX and other trading venues, indicating that the NASDAQ stock market benefited most from Reg NMS. These results are consistent with our expectation that traders are more likely to send orders to the market (i.e., NASDAQ) that offers a fast and greater probability of execution in the post-NMS period than in the pre-NMS period.

Recently, market observers suggest that the SEC should rethink and revise Reg NMS as it led to an increase in high frequency trading and a deterioration of market liquidity and execution quality. Our empirical results are generally consistent with these observations. Our results also support the view of those who opposed Reg NMS that OPR will reduce market liquidity because it reduces the role of NYSE specialists and floor brokers as the liquidity providers of last resort and as information intermediaries.

Appendix A: Order Protection Rule
1. Key Differences in Trade-Through Provisions between the ITS and
the OPR

ITS                                OPR

ITS applies only to exchange-      OPR applies to both exchange-
  listed stocks. ITS does not      listed stocks and NASDAQ
  apply to NASDAQ stocks.          stocks.
  NASDAQ-listed stocks are
  required to meet price
  priority within a market, but
  not across markets.

Trade-throughs are not allowed     Reg NMS differentiates markets
  in all (fast and slow)           (or quotes) into fast and
  markets.                         slow. Electronic markets are
                                   thought of as fast markets,
                                   while manual, floor-based
                                   exchanges are considered slow
                                   markets. The primary purpose
                                   of OPR is to prevent the
                                   execution of trades at prices
                                   inferior to protected
                                   quotations displayed by other
                                   trading centers. Protected
                                   quotation means a protected
                                   bid or a protected offer.

ITS                                OPR
                                   Protected bid or protected
                                   offer means a quotation in an
                                   NMS stock that: 1) is
                                   displayed by an automated
                                   trading center, 2) is
                                   disseminated pursuant to an
                                   effective national market
                                   system plan, and 3) is an
                                   automated quotation that is
                                   the best bid or best offer of
                                   a national securities
                                   exchange, the best bid or best
                                   offer of the NASDAQ Stock
                                   Market, or the best bid or
                                   best offer of a national
                                   securities association other
                                   than the best bid or best
                                   offer of the NASDAQ Stock
                                   Market.

ITS requires order routers to      OPR applies to the
  wait for a response from a       transactions of broker-
  manual market (NYSE).            dealers acting as off-
                                   exchange block positioners in
                                   exchange-listed stocks.

ITS does not apply to the          OPR applies to trade-throughs
  transactions of broker-          of 100-share quotations.
  dealers acting as off exchange
  block positioners in exchange-
  listed stocks.

ITS does not apply to trade-
  throughs of 100-share
  quotations.

2. Arguments For and Against the OPR

Arguments for the OPR             Arguments Against the OPR

Stronger protection of public     OPR is not necessary because
  limit orders (by prohibiting    competition among markets, the
  trade-throughs) would help      duty of best execution, and
  reward liquidity suppliers,     economic self-interest would
  encourages competition among    be sufficient to protect limit
  traders, thus increasing        orders. (Blume, 2002, 2007;
  market liquidity and reducing   O'Hara, 2004).
  trading costs.

Strong intermarket price          OPR may not improve execution
  protection offers greater       quality as it does not
  assurance that investors who    recognize the diversity of
  submit market orders receive    traders and the differential
  the best available prices.      needs of traders. In
                                  particular, OPR prevents
                                  institutional investors from
                                  accessing large amounts of
                                  liquidity at prices that are
                                  slightly worse than the inside
                                  quote, thereby reducing
                                  execution speed and increasing
                                  execution costs. OPR forces
                                  institutional traders to
                                  reveal their intentions by
                                  first executing at the inside
                                  market. For example, a limit
                                  sell order for 100 shares at
                                  the inside market (posted
                                  perhaps by a HFT) must be
                                  executed first rather than,
                                  say, 3,000 shares behind it as
                                  OPR applies only to the top of
                                  the book. This tip of the hand
                                  occurs with no real benefit to
                                  a large buyer as the 100
                                  shares

Arguments for the OPR             Arguments Against the OPR fill
                                  is inconsequential if the
                                  order is for 10,000 shares. As
                                  a result, the liquidity behind
                                  the limit sell order may
                                  disappear immediately forcing
                                  the large order to fill later
                                  at prices that are poorer than
                                  those that were slightly worse
                                  than the inside quote
                                  available at the time of order
                                  submission. (Letter to
                                  Jonathan G. Kats from Bruce N.
                                  Lehmann and Joel Hasbrouck,
                                  Organizers, Regulation NMS
                                  Study Group, 2004).

                                  There is little data in
                                  support of the SEC's view that
                                  trade-throughs impose an
                                  externality on the liquidity
                                  suppliers who are traded
                                  through, thereby reducing
                                  public liquidity. There are
                                  substantial liquidity
                                  provisions on NASDAQ although
                                  it operates without trade-
                                  through prohibitions with
                                  other markets. (Under the ITS,
                                  NASDAQ-listed stocks were
                                  required to meet price
                                  priority within a market but
                                  not across markets.) (Letter
                                  to Jonathan G. Kats from Bruce
                                  N. Lehmann and Joel Hasbrouck,
                                  Organizers, Regulation NMS
                                  Study Group, 2004).

                                  Under the OPR, some traders
                                  would bypass the NYSE for
                                  speedier trades on an
                                  automated system even if the
                                  price available on the
                                  automated system is poorer
                                  than the specialist quote.
                                  Insofar as the order flow
                                  diversion from the NYSE is
                                  large, the new rule would have
                                  a detrimental effect on
                                  liquidity given the finding of
                                  prior research that the NYSE
                                  provides the lowest spread and
                                  price impact (O'Hara, 2004).
                                  OPR would reduce the role of
                                  NYSE specialists as liquidity
                                  providers of last resort
                                  (O'Hara, 2004).

                                  OPR may have a detrimental
                                  effect on price discovery and
                                  market liquidity as it may
                                  increase internalized orders
                                  at the large dealer firms,
                                  removing these orders from a
                                  public market. OPR allows
                                  internalization of order flow
                                  as long as trades occur at or
                                  within the NBBO (O'Hara,
                                  2004).

Appendix B: Access Rule

1. Key Differences between the Pre- and Post-AR

Pre-AR                             Post-AR

Mandates a collective linkage      AR promotes access to
  facility (e.g., ITS) to          quotations through the use of
  facilitate the necessary         private linkages offered by a
  access to quotations.            variety of connectivity
                                   providers. AR requires fair
                                   and nondiscriminatory access
                                   to quotations displayed by SRO
                                   trading centers through
                                   private linkages.

Some trading centers are           AR permits market makers to
  permitted to charge fees and     charge fees for executions of
  some are not. For instance,      orders against their
  although ECNs and other types    quotations, irrespective of
  of trading centers, including    whether the order executions
  SROs, may charge access fees,    are effected on an SRO trading
  market makers are not            facility or directly by the
  permitted to charge any fee      market maker.
  for counterparties accessing
  their quotations. Some trading
  centers have very few fees on
  their books of more than
  $0.003 per share, while others
  earn substantial revenues from
  such fees.

                                   AR establishes an upper bound
                                   ($0.003) on the cost (i.e.,
                                   the access fee) of accessing
                                   such quotations.
                                   AR requires each SRO to adopt,
                                   maintain, and enforce rules
                                   that will prevent their
                                   members from displaying
                                   quotations that lock or cross
                                   the protected quotations of
                                   other trading centers. Trading
                                   centers are allowed, however,
                                   to display automated
                                   quotations that lock or cross
                                   the manual quotations of other
                                   trading centers.

2. Arguments For and Against the AR

Arguments for the AR               Arguments against the AR

AR would level the playing         Competition alone would be
  field in terms of who could      sufficient to address the
  charge fees, as well as give     access fee issue. AR imposes
  greater certainty to market      an arbitrary price ceiling on
  participants that quoted         the transaction and removes
  prices will, essentially, be     market forces from setting
  true prices.                     access fees. The SEC's
                                   approach is both
                                   anticompetitive and
                                   inconsistent with the general
                                   view that market prices are
                                   best set by the market
                                   (O'Hara, 2004).

AR increases the accuracy of       Reg NMS re-establishes the SEC
  displayed quotations by          as a regulator of fees and the
  establishing an upper bound on   regulation of fees will
  the cost (i.e., the access       ultimately be detrimental to
  fee) of accessing such           investors as it reduces
  quotations. The uniform fee      competition. Reg NMS will
  limitation of $0.003 promotes    negate innovative developments
  equal regulation of markets      and deter competition between
  and broker-dealers by applying   markets because it imposes the
  equally to all types of          SEC's views on how the markets
  trading centers and all types    should be structured and how
  of market participants, and      equities should be traded. The
  supports the integrity of the    micromanagement of market
  price protection requirement     structure and trading rules by
  established by OPR. For          the SEC will ultimately be
  instance, although ECNs and      detrimental to investors as it
  other types of trading           reduces competition and
  centers, including SROs, may     inhibits innovation. (Blume,
  charge access fees, market       2002, 2007).
  makers are not permitted to
  charge any fee for
  counterparties accessing their
  quotations. AR permits market
  makers to charge fees for
  executions of orders against
  their quotations, irrespective
  of whether the order
  executions are effected on an
  SRO trading facility or
  directly by the market maker.

AR complements OPR as it helps
  to protect the best displayed
  quotes against trade-through
  by allowing broker-dealers and
  trading centers to access
  those quotes easily and
  inexpensively.

A single, uniform fee
  limitation of $0.003 does not
  materially interfere with
  business practices.

It is consistent with the
  Quote Rule for market makers
  to charge fees for access to
  their quotations, so long as
  such fees meet the
  requirements of AR.

References

Anand, A., A. Puckett, P.J. Irvine and K. Venkataraman, 2010, "Market Crashes and Institutional Trading," Available at SSRN: http://ssrn.com/abstract=1524845.

Barclay, M., T. Hendershott and C. Jones, 2008, "Order Consolidation, Price Efficiency, and Extreme Liquidity Shocks," Journal of Financial and Quantitative Analysis 43, 93-121.

Battalio, R., A. Ellul and R. Jennings, 2007, "Reputation Effects in Trading on the New York Stock Exchange," Journal of Finance 62, 1243-1271.

Benveniste, L., A. Marcus and W. Wilhelm, 1992, "What's Special about the Specialist?" Journal of Financial Economics 32, 61-86.

Bessembinder, H., 2003, "Issues in Assessing Trade Execution Costs," Journal of Financial Markets 6, 233-257.

Bessembinder, H. and H. Kaufman, 1997, "A Comparison of Trade Execution Costs for NYSE and NasdaqListed Stocks," Journal of Financial and Quantitative Analysis 32, 287-310.

Bessembinder, H., W. Maxwell and K. Venkataraman, 2006, "Market Transparency, Liquidity Externalities, and Institutional Trading Costs in Corporate Bonds," Journal of Financial Economics 82, 251-288.

Blume, M., 2002, "The Structure of US Equity Markets," Brookings-Wharton Papers on Financial Services, Washington, D.C, The Brookings Institution.

Blume, M., 2007, "Competition and Fragmentation in the Equity Markets: The Effect of Regulation NMS," Available at SSRN: http://ssrn.com/abstract=959429.

Boehmer, E., 2005, "Dimensions of Execution Quality: Recent Evidence for US Equity Markets," Journal of Financial Economics 78, 463-704.

Bollen, N. and R. Whaley, 1998, "Are 'Teenies' Better?" Journal of Portfolio Management 25, 10-24.

Brunnermeier, M., 2009, "Deciphering the Liquidity and Credit Crunch 2007-2008," Journal of Economic Perspectives 23, 77-100.

Comerton-Forde, C., T. Hendershott, C. Jones, P. Moulton and M. Seasholes, 2010, "Time-Variation in Liquidity: The Role of Market-Maker Inventories and Revenues," Journal of Finance 65, 295-331.

Dai, N., H. Jo and J. Schatzberg, 2010, "The Quality and Price of Investment Banks' Service: Evidence from the PIPE Market," Financial Management 39, 585-612.

Demiroglu, C. and M. Ryngaert, 2010, "The First Analyst Coverage of Neglected Stocks," Financial Management 39, 555-584.

Diether, K., K. Lee and I. Werner, 2009, "It's SHO Time! Short-Sale Price Tests and Market Quality," Journal of Finance 64, 37-73.

Foucault, T. and A. Menkveld, 2008, "Competition for Order Flow and Smart Order Routing Systems," Journal of Finance 63, 119-158.

Harris, L., 1993, "Consolidation, Fragmentation, Segmentation and Regulation," Financial Markets, Institutions & Instruments 2, 1-28.

Hasbrouck, J., 1993, "Assessing the Quality of a Security Market: A New Approach to Transaction-Cost Measurement," Review of Financial Studies 6, 191-212.

Hendelman, S. and B. Rowley, 2010, "A Flawed Model: Relying on High Frequency Traders as Liquidity Providers," Advanced Trading, http://www.advancedtrading.com/algorithms/227500286#comments.

Hendershott, T. and C. Jones, 2005, "Island Goes Dark: Transparency, Fragmentation, and Regulation," Review of Financial Studies 18, 743-793.

Hendershott, Y. and P.C. Moulton, 2011, "Automation, Speed, and Stock Market Quality: The NYSE's Hybrid," Journal of Financial Markets 14, 568-604.

Huang, R. and H. Stoll, 1996, "Dealer versus Auction Markets: A Paired Comparison of Execution Costs on NASDAQ and the NYSE," Journal of Financial Economics 41, 313 -357.

Khandani, A.E. and A. Lo, 2007, "What Happened to the Quants in August 2007?" Available at SSRN: http://ssrn.com/abstract=1015987.

Lee, C. and M. Ready, 1991, "Inferring Trade Direction from Intraday Data," Journal of Finance 46, 733-746.

Macey, J. R. and M. O'Hara, 1999, "Regulating Exchanges and Alternative Trading Systems: A Law and Economics Perspective," Journal of Legal Studies 28, 17-54.

Madhavan, A., 1995, "Consolidation, Fragmentation, and the Disclosure of Trading Information," Review of Financial Studies 8, 579-603.

Madhavan, A., 2000, "Market Microstructure: A Survey," Journal of Financial Markets 3, 205-258.

Mehta, N., 2010, "Missing: The Stock Exchange Buyers of Last Resort," Bloomberg Businessweek, http://www.businessweek.com/magazine/content/10_40/b4197042454014.htm.

O'Hara, M., 2004, "Searching for a New Center: US Securities Markets in Transition," Federal Reserve Bank of Atlanta Economic Review, 37-52.

O'Hara, M. and M. Ye, 2011, "Is Market Fragmentation Harming Market Quality?" Journal of Financial Economics 100, 459-474.

Petersen M.A., 2009, "Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches," Review of Financial Studies 22, 435-480.

Popescu, M. and Z. Xu, 2011, "Co-managers, Information, and the Secondary Market Liquidity of Initial Public Offerings," Financial Management 40, 199-218.

Redler, S., 2010, "SEC Should Rethink Reg NMS to Fix HFT Liquidity Problem," MrSwing, http://www.mrswing.com/articles/ SEC_Should_Rethink_Reg_NMS_to_Fix_HFT_Liquidity_Pr.html.

Zhao, X. and K.H. Chung, 2007, "Information Disclosure and Market Quality: The Effect of SEC Rule 605 on Trading Costs," Journal of Financial and Quantitative Analysis 42, 657-682.

(1) The 1975, Amendments to the Securities Exchange Act of 1934 laid the foundation for creating a National Market System (NMS) in which all equity trades are integrated into a single computerized trading system. On April 6, 2005, the Securities and Exchange Commission (SEC) adopted Reg NMS to establish such a system in the US equity markets.

(2) See Regulation NMS, Exchange Act Release No. 34-51808 (June 9, 2005) ("Adopting Release"). A protected quotation is the top of book quotation displayed by an automated trading center.

(3) An analysis of the SEC staff indicates that more than 12 billion shares of displayed quotations in NASDAQ and NYSE stocks were traded through in 2003 by, on average, $02.3 for NASDAQ stocks and $02.2 for NYSE stocks. These traded-through quotations amount to a total of $321 million in bypassed limit orders and inferior prices for investors.

(4) For example, if two trading centers displayed an identical ask price of $20 for an NMS stock, but their access fees were different (say, $0.001 vs. $0.009), the net cost of buying this stock would be $20.00l in one center and $20.009 in the other.

(5) Others disagree. Opponents of OPR believe that given public availability and easy access to each market's quotations, competition among markets, the duty of best execution, and economic self-interest would be sufficient to protect limit orders and produce the most fair and efficient markets.

(6) Regulation NMS, Exchange Act Release No. 34-51808 (June 9, 2005) ("Adopting Release"). Dissent of Commissioners Cynthia A. Glassman and Paul S. Atkins to the Adoption of Regulation NMS (June 9, 2005), available at http://www.sec.gov/.

(7) Harris (1993) discusses the real world difficulties of facilitating multimarket competition without damaging liquidity externalities (e.g., more efficient prices) that arise when traders come together in space. Liquidity externalities are one of the most important issues in market design (Madhavan, 2000) and market regulation (Macey and O'Hara, 1999).

(8) See a letter to Jonathan G. Kats from Bruce N. Lehmann and Joel Hasbrouck, Organizers, Regulation NMS Study Group, available at http://www.sec.gov/rules/proposed/s71004/10academicstudy052304.pdf.

(9) Hendershott and Moulton (2009) find that the NYSE's Hybrid market, which expanded automated electronic execution and reduced specialist intervention on the NYSE in 2006, led to an increase in the bid-ask spreads. They attribute this result to larger adverse selection costs in a fast, automated execution system.

(10) See http://www.nyxdata.com/factbook.

(11) The NYSE argued strongly against changing the trade-through rules (i.e., allowing fast markets to trade-through better posted prices on slow markets), citing the costs that would arise from executing trades at prices from one to four cents off of the best quote. The NYSE provides empirical estimates ranging from $1.5 billion to $3.5 billion due to this effect alone.

(12) OPR allows internalization of order flow as long as trades occur at or within the NBBO. In addition, Reg NMS facilitates internalization of order flow through the sub-penny rule. According to this rule, broker-dealers can internalize order flow and trade with a market order at smaller increments than a penny. Thus, broker-dealers can trade on their own account at a trivial price improvement when they think it is in their interest. This advantage as well as knowledge of order flow may encourage them to internalize order flow.

(13) Prior studies use these measures as proxies for liquidity (Dai, Jo, and Schatzberg, 2010; Demiroglu and Ryngaert, 2010; Popeseu and Xu, 2011).

(14) See Bollen and Whaley (1998) for this measure.

(15) The long time lapse between rule adoption (April 6, 2005) and rule implementation (July 9, 2007) was largely due to the delay in the implementation of the trade-through rule that was originally scheduled to phase-in June 2006. Trading centers put pressure on the SEC to delay implementation of the trade-through rule to resolve a number of implementation issues. In addition, the SEC gave the floor-based exchanges (NYSE and AMEX) sufficient time to adopt a new automated market structure.

(16) Data source: http://www.nyxdata.com/factbook.

(17) We report the results for NYSE, AMEX, and NASDAQ stocks as a whole. The results are qualitatively similar when we conduct our analysis using the separate samples of NYSE/AMEX stocks and NASDAQ stocks.

(18) Although the NYSE's hybrid system was implemented before Reg NMS, the former might have been a direct response to the latter if market participants had responded to key elements of Reg NMS before its official implementation date. In such a case, an accurate measurement of the net effect of Reg NMS on market quality would require an assessment of the effect of the NYSE's new architecture on market quality.

(19) Our results are qualitatively identical when we extend the post-NMS period to October 8, 2007, which is the phase-in completion date for the main group.

(20) Bessembinder, Maxwell, Venkataraman (2006) find that the initiation of the National Association of Securities Dealer's Trade Reporting and Compliance Engine (TRACE) led to a reduction in execution costs not only for bonds eligible for TRACE reporting, but also for bonds not eligible for TRACE reporting. Similarly, Zhao and Chung (2007) analyze the effect of posttrade transparency on market quality and find evidence that market centers began to post uniformly narrower spreads for all NYSE stocks (not just for the first phase-in stocks) when they were required to report execution quality.

(21) The increase in return volatility may be explained, at least in part, by the increased uncertainty in the economy due to the credit crunch and quant meltdown (see Section IV).

(22) The primary objective of Reg NMS is to minimize investor transaction costs. See Regulation NMS, Exchange Act Release No. 34-51808 (June 9, 2005) ("Adopting Release"), page 11.

(23) See Hendelman and Rowley (2010) and Mehta (2010). Others had expressed concerns on this issue due to the fact that OPR prohibits institutional investors from accessing large amounts of liquidity at prices slightly worse than the inside quote. See a letter to Jonathan G. Kats from Bruce N. Lehmann and Joel Hasbrouck, Organizers, Regulation NMS Study Group, available at http://www.sec.gov/rules/proposed/s71004/10academicstudy052304.pdf.

(24) During the Flash Crash (on May 6, 2010, the US equity markets experienced a dramatic fall and subsequent recovery in what has since become known as the "Flash Crash"), several major HFT firms shut down their systems to protect themselves and did not provide any liquidity to the market.

(25) See Brurmermeier (2009) for a detailed analysis of these events.

(26) Anand et al. (2010) examine institutional trading in US equities during the financial crisis of 2007-2008 and find that institutional traders experienced a dramatic increase in trading costs and higher uncertainty in execution prices during the crisis.

(27) Diether, Lee, and Werner (2009) find wider spreads when the uptick rule is suspended for 1,000 of the Russell 3,000 stocks in 2005. The same effect may be observed for the remaining stocks when the uptick rule is completely repealed in 2007.

(28) This method requires that pilot stocks were chosen randomly or for reasons unrelated to their liquidity.

(29) The residuals of a given firm may be correlated across days (time-series dependence) and/or the residuals of a given day may be correlated across different firms (cross-sectional dependence).

(30) We used the SAS code available at https://webspace.utexas.edu/johnmac/www/cluster%20code.htm.

(31) All other results are available from the authors upon request.

(32) Because the pricing error has zero mean, its volatility measures the magnitude of the pricing error as well.

(33) Boehmer (2005) analyzed market order execution quality using the SEC Rule 605 data and found that inferring execution quality from costs alone is problematic.

(34) See a letter to Jonathan G. Kats from Bruce N. Lehmann and Joel Hasbrouck, Organizers, Regulation NMS Study Group, available at http://www.sec.gov/rules/proposed/s71004/10academicstudy052304.pdf.

(35) We find qualitatively similar results for the pilot group. The results are available from the authors upon request.

(36) Although the results from Rule 605 data are qualitatively similar to those from the TAQ database, actual figures from the two databases are somewhat different for a couple of reasons. For example, Rule 605 effective spreads are share-weighted, whereas TAQ effective spreads are trade-weighted and stocks included in our Rule 605 data are not exactly identical to those included in the TAQ data. Traders Magazine (see http://www.tradersmagazine.com/issues/ 20_278/100330-1.html?zkPrintable=true) reported that according to an analysis conducted by SEC economists, both spreads and depths improved since Reg NMS went into effect. These results are not directly comparable to our results as SEC economists measured the effect of Reg NMS on market quality by comparing Rule 605 data from April 2005 to April 2007. Our pre- and post-NMS periods for the 605 execution quality data are May/June 2007 and September/October 2007, respectively.

(37) Rule 605 statistics include all orders sent to a market center including fast and slow executions.

(38) Boehmer (2005) finds that NASDAQ provides faster executions than the NYSE for market orders that are smaller than 2,000 shares.

(39) NASDAQ provided better executions than the NYSE in terms of smaller effective spreads, smaller price impact, a greater proportion of trades that received price improvement, a lower proportion of trades that are executed outside the quote, faster execution speed, higher fill rates, and lower order cancellation rates.

Kee H. Chung and Chairat Chuwonganant *

The paper benefited greatly from the comments and suggestions of an anonymous referee and Bill Christie (editor). The authors thank Benjamin Blau, Joon Chae, Jae M. Chung, Bong-Chan Kho, Beom-Sik Jang, Dongcheol Kim, Jung-Wook Kim, Kuan-Hui Lee, Mingsheng Li, Thomas McInish, Carl Shen, Matthew Spiegel, James Upson, Hao Zhang, Xin Zhao, Lingyin Zhu, colleagues at SUNY-Buffalo and Kansas State University, session participants at the 2010 Financial Management Association Conference, and seminar participants at Seoul National University and the Korea University Business School for valuable discussion, comments, and suggestions. The usual disclaimer applies.

* Kee H. Chung is the Louis M. Jacobs Professor of Financial Planning and Control in the School of Management at the State University of New York (SUNY) at Buffalo in Buffalo. NY. Chairat Chuwonganant is an Associate Professor of Finance in the College of Business Administration at Kansas State University in Manhattan, KS.

Table I. Comparison of Variables between the Pre- and Post-Reg NMS
Periods

This table compares the measures of spreads, depth, market quality
index, volatility, price impact, number of trades, and dollar trade
size between the pre-and post-NMS periods for the pilot and main
stocks. Quoted dollar spread is the difference between the ask price
and bid price. Quoted percentage spread is the quoted dollar spread
divided by the quote midpoint. Effective dollar spread is defined as
[2D.sub.i,t] ([P.sub.i,t] -[M.sub.i,t]), where [P.sub.i,t] is the
transaction price of the stock at time t, [M.sub.i,t] is the quote
midpoint of the stock at time t, and [D.sub.i,t] is a binary variable
equal to +1 for customer buy orders and -1 for customer sell orders.
Effective percentage spread is the effective dollar spread divided by
the quote midpoint. Depth is the sum of the dollar bid size and the
ask size. Market quality index is defined as half the sum of the
dollar bid size and the ask size divided by the quoted percentage
spread. Volatility is the standard deviation of daily returns. Price
Impact is defined as [D.sub.i,t] ([M.sub.i,t +5] -[M.sub.i,t],), where
[M.sub.i,t] is the quote midpoint of the stock at time t and
[M.sub.i,t] + 5 is the quote midpoint of the stock at t + five
minutes. Number of trades is the daily number of trades. Dollar trade
size is the dollar transaction size. For the pilot stocks, the pre-
NMS period is from May 24, 2007 to July 6, 2007 and the post-NMS
period is from July 9, 2007 to August 17, 2007. For the main stocks,
the pre-NMS period is from July 9, 2007 to August 17, 2007 and the
post-NMS period is from August 20, 2007 to October 1, 2007. We
calculate the mean values of each variable for each stock during the
periods prior to and after the implementation of the Reg NMS. % [DELTA]
represents the percentage differences in values between the pre and
post periods ([postvalue -prevalue]-prevalue]). N denotes the number
of observations. The numbers in parentheses are t-statistics.

                        Pilot Group (N = 242)

                        Pre      Post     Post - Pre    (% [DELTA])

Panel A. Spreads, Depths, and Market Quality Index

Quoted spread ($)       0.0237   0.0299   0.0062 ***    (0.2616)
                                          (10.83)
Quoted spread (%)       0.0019   0.0023   0.0004 ***    (0.2105)
                                          (8.67)
Effective spread ($)    0.0198   0.0263   0.0065 ***    (0.3283)
                                          (16.52)
Effective spread (%)    0.0014   0.0018   0.0004 ***    (0.2857)
                                          (12.41)
Quoted depth ($)        81,618   55,601   -26,017 ***   (-0.3188)
                                          (-9.16)
Market quality          2.3454   1.4791   -0.8663 ***   (-0.3694)
index (x [10.sup.6])                      (-8.80)

Panel B. Other stock attributes]

Volatility              0.0010   0.0013                 (0.3000)
                                          (10.74)
Price impact ($)        0.0132   0.0180   0.0048 ***    (0.3636)
                                          (9.46)
Number of trades        7,786    10,707   2,921 ***     (0.3751)
                                          (6.88)
Dollar trade size ($)   10,236   8,670    -1,566 ***    (-0.1530)
                                          (-8.32)

                        Main Group (N = 5,937)

                        Pre      Post     Post - Pre    (% [DELTA])

Panel A. Spreads, Depths, and Market Quality Index

Quoted spread ($)       0.0611   0.0684   0.0073 ***    (0.1195)
                                          (14.95)
Quoted spread (%)       0.0043   0.0051   0.0008 ***    (0.1860)
                                          (9.56)
Effective spread ($)    0.0429   0.0482   0.0053 ***    (0.1235)
                                          (24.55)
Effective spread (%)    0.0027   0.0032   0.0005 ***    (0.1852)
                                          (42.34)
Quoted depth ($)        42,715   38,140   -4,575 ***    (-0.1071)
                                          (-11.60)
Market quality          0.7109   0.6066   -0.1043 ***   (-0.1467)
index (x [10.sup.6])                      (-12.50)

Panel B. Other stock attributes]

Volatility              0.0022   0.0026   0.0004 ***    (0.1818)
                                          (16.07)
Price impact ($)        0.0292   0.0319   0.0027 ***    -0.0925
                                          (13.16)       9
Number of trades        3,404    3,686    282 ***       (0.0828)
                                          (9.48)
Dollar trade size ($)   6,934    6,387    -547 ***      (-0.0789)
                                          (-12.92)
*** Significant at the 0.01 level.

Table II. Regression Results Using Changes in Variables between the
Pre- and Post-NMS Periods

To examine whether the changes in the spread, depth, and market
quality index are indeed due to Reg NMS after controlling for changes
in the market environment, we estimate the following regression model
for the pilot and main groups of stocks: [VAR.sub.i.sup.post] -
[VAR.sub.i.sup.pre] = [[beta].sub.O] + [SIGMA][[beta].sub.k]
([X.sup.post.sub.ki] -[X.sup.pre.sub.ki]) + [[epsilon].sub.i];; where
superscripts "post" and "pre" denote the post-and pre-NMS periods,
respectively, VAR; denotes the quoted spread, effective spread, dollar
depth, or market quality index of stock i, [X.sub.k] (k = 1 to 4)
represents one of the four stock attributes ([PRICE.sub.i],
[VOLUME.sub.i], [VOLA.sub.i], or [RET.sub.i]). PRICE is the mean share
price, VOLUME is the mean daily dollar trading volume, VOLA is the
standard deviation of daily returns, and RET is the average return
during the period. The intercept [[beta].sub.o] measures the
difference in variables between the pre-and post-NMS periods after
controlling for changes in the four stock attributes. For pilot stocks,
the pre-NMS period is from May, 24, 2007 to July 6, 2007 and the
post-NMS period is from July 9, 2007 to August 17, 2007. For the main
group of stocks, the pre-NMS period is from July 9, 2007 to
August 17, 2007 and the post-NMS period is from August 20, 2007 to
October 1, 2007. Panel A reports the results for the pilot group,
while Panel B presents the results for the main group. N denotes the
number of observations. Numbers in parentheses are t-statistics.

               [DELTA]       [DELTA]       [DELTA]        [DELTA]
               Quoted        Quoted        Effective      Effective
               Spread ($)    Spread (%)    Spread ($)     Spread (%)

Panel A. Regression Results for the Pilot Group (N= 242)

Intercept      0.0064 ***    0.0004 ***     0.0063 ***    0.0004 ***
               (13.19)       (9.67)         (12.56)       (9.25)
[DELTA]PRICE   0.0005 ***    -0.0001 ***    0.0003 ***    -0.0001 **
               (6.64)        (-2.93)        (3.31)        (-2.49)
[DELTA]        -0.0105 ***   -0.0011 ***    -0.0091 ***   -0.0011 ***
VOLUME         (-2.82)       (-3.23)        (-2.70)       (-4.97)

[DELTA]RET     0.2709 ***    0.0140 **      0.1941 ***    0.0102 **
               (3.68)        (2.17)         (3.06)        (2.41)
[DELTA]VOLA    0.0922 ***    0.0094 ***     0.1022 ***    0.0091 ***
               (2.75)        (3.19)         (3.54)        (4.72)
[R.sup.2]      0.12          0.10           0.11          0.18

Panel B. Regression Results for the Main Group (N= 5,937)

Intercept      0.0074 ***     0.0008 ***    0.0051 ***    0.0005 ***
               (21.79)       (18.56)        (26.93)       (37.07)
[DELTA]PRICE   0.0008 ***     -0.0001 ***   0.0009 ***    -0.0001 ***
               (9.69)        (-5.15)        (18.52)       (-4.78)
[DELTA]        0.0181 ***    -0.0017 **     -0.0110 ***   -0.0004 **
VOLUME         (-3.11)       (-2.35)        (-3.22)       (-2.29)

[DELTA]RET     0.0966 **      0.0044 **     0.0552 **     0.0035 ***
               (2.48)        (2.10)         (2.42)        (2.99)
[DELTA]VOLA    0.1017 ***     0.0154 ***    0.0459 ***    0.0026 ***
               (4.46)        (5.97)         (3.44)        (3.84)
[R.sup.2]      0.07          0.05           0.09          0.04

                                 [DELTA]
                                 Market
               [DELTA]           Quality
               Depth             Index
               ($ in millions)   (x [10.sup.-6])

Panel A. Regression Results for the Pilot Group (N= 242)

Intercept      -0.0262 ***       -0.8629 ***
               (-8.35)           (-8.81)
[DELTA]PRICE   -0.0017 **        -0.0880
               (-2.30)           (-3.41)
[DELTA]        0.0661 ***        1.8878 **
VOLUME         (2.67)            (2.35)

[DELTA]RET     -1.1943 **        -48.4825 ***
               (-2.44)           (-2.79)
[DELTA]VOLA    -0.4992 **        -17.8006 **
               (-2.24)           (-2.15)
[R.sup.2]      0.07              0.08

Panel B. Regression Results for the Main Group (N= 5,937)

Intercept      -0.0045 ***       -0.1046 ***
               (-9.53)           (-9.32)
[DELTA]PRICE   -0.0002 **        -0.0066
               (-2.19)           (-2.73)
[DELTA]        0.0176 **          0.4099 **
VOLUME         (2.4l)            (2.40)

[DELTA]RET     -0.1002 **        -3.2757 ***
               (-2.05)           (-2.87)
[DELTA]VOLA    -0.0652 **        -1.5626 **
               (-2.28)           (-2.34)
[R.sup.2]      0.03              0.03

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

Table III. Stock Attributes of Pilot and Control Stocks

The table presents the stock attributes of our pilot and control
sample stocks. To obtain the control sample of  NYSE pilot stocks, we
first calculate the CMS of each pilot stock against each and every
NYSE-listed security  with the same two-digit SIC code in the main
implementation group: CMS = [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE
IN ASCII.], where [X.sub.k] represents one of the four stock
attributes (i.e., price, trading volume, return, and  return
volatility) and [SIGMA] denotes the summation over k = 1 to 4. Then,
for each pilot stock, we select the  stock in the main group with the
lowest score. Once we match a main stock on a pilot issue, we no
longer consider that particular main stock for subsequent matches. We
obtain the AMEX and NASDAQ control stocks in the same manner. We use
the stock attributes in the pre-NMS period for matching. N denotes the
number of observations. Numbers in parentheses are t-statistics.

                          Pilot      Control     Pilot--
                         (N = 242)   (N = 242)   Control

Share price ($)           33.08       32.80       0.28(0.35)
Volume ($ in millions)   124.08      122.61       1.47(0.57)
Return                    -0.0013     -0.0014     0.0001(0.28)
Volatility                 0.0010      0.0011    -0.0001(-0.39)

Table IV. Difference-in-Difference Regression Results

Panel A provides the results of the following regression model using
the pilot and control (matching) sample of stocks: [MATHEMATICAL
EXPRESSION NOT REPRODUCIBLE IN ASCII.]; were [VAR.sub.i,t] is the
quoted dollar spread, quoted percentage spread, effective dollar
spread, effective percentage spread, dollar depth, or market quality
index of stock i on day t, [D.sub.t.sup.NMS] is an indicator variable
equal to one for post-NMS days (i.e., July 9, 2007 to August 17,
2007) and zero for pre-NMS days (i.e., May 24, 2007 to July 6, 2007),
[X.sub.i,t,k] (k = 1 to 4) represents one of the four stock attributes
(PRICE, RET, VOLUME, and VOLA) of stock i on day t, and [SIGMA]
denotes the summation over k = 1 to 4. PRICE is the mean share price,
RET is the mean return, VOLUME is the mean daily dollar trading
volume, and VOLA is the standard deviation of quote midpoint returns.
The superscripts "pilot" and "control" represent the pilot and control
stocks, respectively. We include the matched pair fixed effect
([[lambda].sub.i]) to control for any differences between two stocks
in a pair that are present during the pre-NMS period. We include the
time dummy variable ([[lambda].sub.t]) to control for any market-wide
changes in the dependent variable. Panel B reports the results of the
following regression model using the main and control sample of stocks:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.]; where
[D.sub.t.sup.NMS] is an indicator variable equal to one for post-NMS
days (i.e., August 20, 2007 to October 1, 2007) and zero for pre-NMS
days (i.e., July 9, 2007 to August 17, 2007), the superscripts "main"
and "control" represent the main and control stocks, respectively,
and all other variables are the same as defined above. N denotes the
number of observations. Numbers in parentheses are t-statistics.

                   [DELTA]Quoted   [DELTA]Quoted    [DELTA]Effective
                      Spread           Spread             Spread
                        ($)             (%)                ($)

Panel A. Regression Results for the Pilot Group (N= 14,520)

[D.sup.NMS]           0.0061 ***       0.0003 ***         0.0059 ***
                     (4.59)           (3.92)             (6.73)
[DELTA]PRICE ($)      0.0003 ***      -0.0001 ***         0.0002 ***
                    (15.59)          (-4.52)            (20.21)
[DELTA]VOLUME        -0.0087 ***      -0.0003 ***        -0.0133 ***
($ in billions)    (-10.80)          (-3.70)            (-3.15)
[DELTA]RET           -0.0108 **       -0.0011 ***        -0.0079 **
                    (-2.27)          (-2.83)            (-2.06)
[DELTA]VOLA           0.0102 ***       0.0014 ***         0.0162 ***
                     (8.90)          (16.04)            (18.36)
[R.sup.2]             0.45             0.39               0.41

Panel B. Regression Results for the Main Group (N = 14,520)

[D.sup.NMS]           0.0070 ***       0.0007 ***         0.0048 ***
                     (5.29)           (4.89)             (4.35)
[DELTA]PRICE ($)      0.0004 ***      -0.0001 ***         0.0004 ***
                    (10.56)          (-5.82)            (17.43)
[DELTA]VOLUME        -0.0093 ***      -0.0006 ***        -0.0027 ***
($ inbillions)      (-3.69)          (-2.93)            (-2.12)
[DELTA]RET           -0.0079          -0.0004 **         -0.0102 **
                    (-1.82)          (-2.32)            (-2.45)
[DELTA]VOLA           0.0395 ***       0.0043 ***         0.0376 ***
                    (13.39)          (26.15)            (22.44)
[R.sup.2]             0.38             0.34               0.37

                   [DELTA]Effective   [DELTA]Depth      [DELTA]Market
                      Spread          ($ in millions)       Quality
                        (%)                                  Index
                                                        (x [10.sup.-6])

Panel A. Regression Results for the Pilot Group (N= 14,520)

[D.sup.NMS]          0.0003 ***        -0.0249 ***       -0.8572 ***
                    (4.48)            (-5.88)           (-4.98)
[DELTA]PRICE ($)    -0.0001 ***        -0.0004 ***       -0.0192
                   (-5.50)            (-7.45)           (-4.54)
[DELTA]VOLUME       -0.0002 ***         0.0181 ***        0.9789 ***
($ in billions)    (-3.26)            (28.75)            (5.74)
[DELTA]RET          -0.0007 ***         0.0394 ***        0.0187 ***
                   (-3.51)             (2.82)            (2.95)
[DELTA]VOLA          0.0017 ***         0.0115 ***       -0.3091 ***
                   (32.22)            (-2.77)           (-2.64)
[R.sup.2]            0.44               0.59              0.35

Panel B. Regression Results for the Main Group (N = 14,520)

[D.sup.NMS]          0.0004 ***        -0.0042 ***       -0.0992 ***
                    (3.97)            (-5.43)           (-4.73)
[DELTA]PRICE ($)    -0.0001 ***        -0.0002 **        -0.0458 ***
                   (-7.26)            (-2.48)           (-6.35)
[DELTA]VOLUME       -0.0002 ***         0.0367 ***        2.0880
($ inbillions)     (-2.79)            (13.56)           (16.05)
[DELTA]RET          -0.0007 ***         0.0262 **         2.7257 ***
                   (-3.40)             (2.16)            (3.32)
[DELTA]VOLA          0.0038 ***        -0.0121 ***       -0.5617 **
                   (33.63)            (-3.34)           (-2.38)
[R.sup.2]            0.39               0.51              0.32

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

Table V. Comparison of Hasbrouck's Pricing Error Measure between the
Pre- and Post-NMS Periods for the Matched Sample

This table reports the Hasbrouck's pricing error measure of NYSE-
AMEX-and NASDAQ-listed stocks for the test group and the control
sample. For the pilot stocks, the pre period (pre) is from May 24,
2007 to July 6, 2007 and the post period (post) is from July 9, 2007
to August 17, 2007. For the main implementation group of stocks, the
pre-NMS period is from July 9, 2007 to August 17, 2007 and the post-
NMS period is from August 20, 2007 to October 1, 2007. Panel A
presents the results for the pilot group, while Panel B provides the
results for the main implementation group. We indicate whether the
difference in Hasbrouck's pricing error measure between the pre-and
post-NMS periods is statistically significant. We also show whether
the difference in Hasbrouck's pricing error measure between the pre-
and post-NMS periods for the test group is statistically different
from the difference in the pricing error measure between the pre-and
post-NMS periods for the control sample. N denotes the number of
observations. Numbers in parentheses are t-statistics.

                           NYSE/AMEX (N = 146)

                  Pre      Post     Post-Pre

Panel A. Results for the Pilot Group of Stocks

Pilot group       0.0056   0.0063   0.0007 ***   (4.81)
Control sample    0.0053   0.0055   0.0002       (0.61)
(Pilot-control)                     0.0005 ***   (4.47)

Panel B. Results for the Main Implementation Group of Stocks

Main group        0.0059   0.0096   0.0037 ***   (8.17)
Control sample    0.0063   0.0065   0.0002       (1.43)
(Main-control)                      0.0035 ***   (5.49)

                           NASDAQ (N = 96)

                  Pre      Post     Post-Pre

Panel A. Results for the Pilot Group of Stocks

Pilot group       0.0041   0.0049   0.0008 ***   (5.27)
Control sample    0.0043   0.0044   0.0001       (0.29)
(Pilot-control)                     0.0007 ***   (4.92)

Panel B. Results for the Main Implementation Group of Stocks

Main group        0.0042   0.0073   0.0031 ***   (9.02)
Control sample    0.0049   0.0052   0.0003       (1.36)
(Main-control)                      0.0028 ***   (5.83)

*** Significant at the 0.01 level.

Table VI. Comparison of SEC Rule 605 Execution Quality Measures for
the Main Group in this table, we compare execution quality measures
between the pre- and post-NMS periods for our main sample stocks by
order type. We also compare the NYSE/AMEX and NASDAQ execution quality
for NYSE/AMEX main stocks in the pre- and post-NMS periods. The table
reports the following execution quality measures: 1) the effective
spread for executions of covered orders, 2) the price impact for
executions of covered orders, 3) the proportion of shares that are
executed at the quote, with price improvement, and outside the quote,
respectively, 4) the share-weighted average duration of time in seconds
from the time of order receipts to the time of order execution for
shares executed at the quote with price improvement and outside the
quote, respectively, and 5) the proportion of shares that are executed
at the receiving market center (Fill rate), executed at other venues
(Away rate), and cancelled prior to execution (Cancelled). We calculate
the mean values of each variable for each stock during the periods
prior to and after the implementation of the Reg NMS. Panel A compares
the execution quality by order types between the pre- and post-NMS
periods. Panel B compares the NYSE/AMEX and NASDAQ execution quality
measures for NYSE/AMEX stocks in the pre- and post-NMS periods. N
denotes the number of observations. Numbers in parentheses are
t-statistics. Panel A. Comparison of Execution Quality between the
Pre- and Post-NMS Periods

                             Pre      Post          Post-Pre
                                      Market orders (N = 3,180)

Effective dollar spread      0.0474   0.0583    0.0109 ***     (30.64)
Price impact                 0.0278   0.0355    0.0077 ***     (17.87)
Proportion of shares that
  are executed
  At the quote               0.5619   0.5169   -0.0450 ***    (-21.17)
  With price improvement     0.0383   0.0702    0.0319 ***     (22.85)
  Outside the quote          0.3957   0.4026    0.0069 ***      (5.05)
Mean execution time for
  shares that are executed
  At the quote               0.2665   1.0187    0.7522 ***      (5.18)
  With price improvement     6.5680   4.4935   -2.0745 ***    (-10.02)
  Outside the quote          1.5752   2.2217    0.6465 ***      (4.02)
Fill rate                    0.7961   0.7376   -0.0585 ***    (-24.13)
Away rate                    0.2003   0.2599    0.0596 ***     (24.66)
Cancelled                    0.0036   0.0025   -0.0011 ***    (-4.26)

                             Nonmarketable limit orders (N = 5,937)

Fill rate                    0.2484   0.2101   -0.0383 ***    (-27.96)
Away rate                    0.0294   0.0168   -0.0126 ***    (-31.75)
Cancelled                    0.7221   0.7731    0.0510 ***     (33.15)

                             Pre       Post      Post - Pre
                             Marketable limit orders (N = 5,937)

Effective dollar spread       0.0379    0.0486    0.0107 ***
Price impact                  0.0274    0.0353    0.0079 ***
Proportion of shares that
  are executed
  At the quote                0.5455    0.5401   -0.0054 ***
  With price improvement      0.0471    0.0525    0.0054 ***
  Outside the quote           0.0625    0.0527   -0.0098 ***
Mean execution time for
  shares that are executed
  At the quote               15.7573   18.1404    2.3831 ***
  With price improvement      2.6707    1.8219   -0.8488 ***
  Outside the quote          18.3656   24.4620    6.0964 ***
Fill rate                     0.5467    0.5445   -0.0022 ***
Away rate                     0.1117    0.1026   -0.0091 ***
Cancelled                     0.3416    0.3527    0.0111 ***

Fill rate
Away rate
Cancelled

                             Post-Pre

Effective dollar spread       (29.27)
Price impact                  (26.55)
Proportion of shares that
  are executed
  At the quote                (-4.90)
  With price improvement      (10.27)
  Outside the quote          (-12.30)
Mean execution time for
  shares that are executed
  At the quote                 (9.77)
  With price improvement      (-6.93)
  Outside the quote            (8.86)
Fill rate                     (-3.14)
Away rate                    (-11.73)
Cancelled                      (9.21)

Fill rate
Away rate
Cancelled

Panel B. Comparison of the Execution Quality of NYSE/AMEX Stocks
between Trades Occurred on the NYSE/AMEX and Trades Occurred on
NASDAQ

                             Execution Quality of NYSE/AMEX Stocks on

                                    Pre-NMS Period

                             NYSE/                NASDAQ-
                             AMEX      NASDAQ      NYSE
Marketable limit orders
(N = 3,180)
Effective dollar spread       0.0252   0.0432     0.0180 ***
Price impact                  0.0212   0.0336     0.0124 ***
Proportion of shares
  that are executed
  At the quote                0.4670   0.5045     0.0375 ***
  With price improvement      0.0092   0.0240     0.0148 ***
  Outside the quote           0.0845   0.0356    -0.0489 ***
Mean execution time for
  shares that are executed
  At the quote               24.8680   6.9049   -17.9631 ***
  With price improvement      2.4212   0.5021    -1.9191 ***
  Outside the quote          26.9930   3.8735   -23.1195 ***
Fill rate                     0.4800   0.4639    -0.0161 ***
Away rate                     0.0812   0.1391     0.0579 ***
Cancelled                     0.4388   0.3970    -0.0418 ***
Nonmarketable limit
  orders (N = 3,180)
Fill rate                     0.2776   0.2421    -0.0355 ***
Away rate                     0.0301   0.0396     0.0095 ***
Cancelled                     0.6923   0.7113     0.0190 ***

                             Execution Quality of NYSE/AMEX Stocks on

                              Pre-NMS
                              Period    Post-NMS Period

                              NASDAQ-    NYSE/
                               NYSE      AMEX     NASDAQ
Marketable limit orders
(N = 3,180)
Effective dollar spread       (31.06)   0.0411    0.0401
Price impact                  (13.62)   0.0327    0.0312
Proportion of shares
  that are executed
  At the quote                (16.70)   0.4831    0.5329
  With price improvement      (29.52)   0.0128    0.0304
  Outside the quote          (-38.73)   0.0686    0.0204
Mean execution time for
  shares that are executed
  At the quote               (-20.16)   27.9230   2.9136
  With price improvement     (-16.37)   2.9184    0.2651
  Outside the quote          (-26.51)   42.2410   3.0645
Fill rate                     (-3.91)   0.4681    0.4756
Away rate                     (20.61)   0.0894    0.1181
Cancelled                    (-14.07)   0.4425    0.4063
Nonmarketable limit
  orders (N = 3,180)
Fill rate                     (-7.11)   0.2122    0.2496
Away rate                      (3.02)   0.0181    0.0345
Cancelled                      (5.92)   0.7697    0.7159

                             Execution Quality of NYSE/AMEX Stocks on

                                Post-NMS Period

                                    NASDAQ-
                                     NYSE
Marketable limit orders
(N = 3,180)
Effective dollar spread       -0.0010 ***    (-4.24)
Price impact                  -0.0015 ***    (-3.82)
Proportion of shares
  that are executed
  At the quote                 0.0498 ***    (19.77)
  With price improvement       0.0176 ***    (32.56)
  Outside the quote           -0.0482 ***   (-68.37)
Mean execution time for
  shares that are executed
  At the quote               -25.0094 ***   (-35.15)
  With price improvement      -2.6533 ***   (-17.99)
  Outside the quote          -39.1765 ***   (-32.91)
Fill rate                      0.0075 ***     (3.34)
Away rate                      0.0287 ***    (12.06)
Cancelled                     -0.0362 ***    (-9.19)
Nonmarketable limit
  orders (N = 3,180)
Fill rate                      0.0374 ***    (20.70)
Away rate                      0.0164 ***    (20.24)
Cancelled                     -0.0538 ***   (-25.79)

                             Execution Quality of NYSE/AMEX Stocks on

                             [NASDAQ.sub.Post] -
                             [NASDAQ.sub.Pre]
Marketable limit orders
(N = 3,180)
Effective dollar spread      -0.0031 ***    (-6.85
Price impact                 -0.0024 ***    (-5.29)
Proportion of shares
  that are executed
  At the quote                0.0284 ***     (7.67)
  With price improvement      0.0064 ***     (6.10)
  Outside the quote          -0.0152 ***    (-5.73)
Mean execution time for
  shares that are executed
  At the quote               -3.9913 ***    (-8.06)
  With price improvement     -0.2370 ***    (-7.46
  Outside the quote          -0.0890 ***    (-7.90)
Fill rate                     0.0117 ***    (16.26)
Away rate                    -0.0210 ***    (-5.89)
Cancelled                     0.0093 ***     (4.97)
Nonmarketable limit
  orders (N = 3,180)
Fill rate                     0.0075 ***     (7.13)
Away rate                    -0.0051 ***    (-5.72)
Cancelled                     0.0046 ***     (4.51)

*** Significant at the 0.01 level.

Table VII. Comparison of Market Shares between the Pre- and Post-NMS
Periods

This table presents the mean proportion of NYSE-AMEX-listed stocks'
trades that are executed on the NYSE-AMEX, NASDAQ, and other trading
venues for the test group and control sample. Likewise, it reports
the mean proportion of NASDAQ-listed stock trades that are executed on
NASDAQ and other trading venues for the test group and control
sample. For the pilot stocks, the pre period (pre) is from May 24,
2007 to July 6, 2007 and the post period (post) is from July 9, 2007
to August 17, 2007. For the main implementation group of stocks, the
pre-NMS period is from July 9, 2007 to August 17, 2007 and the post-
NMS period is from August 20, 2007 to October 1, 2007. Panel A
provides the results for the pilot group, while Panel B presents the
results for the main implementation group. As in other tables, we
show whether the difference in the proportion between the pre-and
post-NMS periods is statistically significant. More importantly, we
also show whether the difference in the proportion between the pre-
and post-NMS periods for the test group is statistically different
from the difference in the proportion between the pre-and post-NMS
periods for the control sample. N denotes the number of observations.
Numbers in parentheses are t-statistics.

                               NYSE/AMEX              NASDAQ

                      Pre      Post     Post-Pre      Pre

Panel A. Results for the Pilot Group of Stocks

NYSE/AMEX (N = 146)
Pilot group           0.4592   0.4121   -0.0471 ***   (-9.94)
Control sample        0.4624   0.4520   -0.0104       (-1.39)
(Pilot-control)                         -0.0367 ***   (-6.29)
NASDAQ (N = 96)
Pilot group
Control sample
(Pilot-control)

Panel B. Results for the Main Implementation Group of Stocks

NYSE/AMEX (N = 146)
Main group            0.4466   0.4105   -0.0361 ***   (-6.06)
Control sample        0.4121   0.3988   -0.0133       (-1.47)
(Main-control)                          -0.0228 ***   (-4.51)
NASDAQ (N = 96)
Main group
Control sample
(Main-control)

                               NASDAQ

                      Pre      Post     Post-Pre

Panel A. Results for the Pilot Group of Stocks

NYSE/AMEX (N = 146)
Pilot group           0.3945   0.4118   0.0173 ***   (5.26)
Control sample        0.3900   0.3958   0.0058       (0.71)
(Pilot-control)                         0.0115 ***   (3.79)
NASDAQ (N = 96)
Pilot group           0.8224   0.8336   0.0112 ***   (4.24)
Control sample        0.8313   0.8334   0.0021       (0.37)
(Pilot-control)                         0.0091 ***   (3.09)

Panel B. Results for the Main Implementation Group of Stocks

NYSE/AMEX (N = 146)
Main group            0.3982   0.4195   0.0213 ***   (5.68)
Control sample        0.4118   0.4208   0.0090       (1.62)
(Main-control)                          0.0123 ***   (3.78)
NASDAQ (N = 96)
Main group            0.8297   0.8494   0.0197 ***   (4.73)
Control sample        0.8336   0.8405   0.0069       (1.54)
(Main-control)                          0.0128 ***   (3.93)

                                   Others

                      Pre      Post     Post-Pre

Panel A. Results for the Pilot Group of Stocks
NYSE/AMEX (N = 146)
Pilot group           0.1464   0.1762    0.0298 ***    (8.11)
Control sample        0.1476   0.1522    0.0046        (0.85)
(Pilot-control)                          0.0252 ***    (5.02)
NASDAQ (N = 96)
Pilot group           0.1765   0.1653   -0.0112 ***   (-4.24)
Control sample        0.1697   0.1676   -0.0021       (-0.37)
(Pilot-control)                         -0.0091 ***   (-3.09)

Panel B. Results for the Main Implementation Group of Stocks

NYSE/AMEX (N = 146)
Main group            0.1550   0.1698    0.0148 ***    (3.43)
Control sample        0.1762   0.1805    0.0043        (0.70)
(Main-control)                           0.0105 ***    (3.27)
NASDAQ (N = 96)
Main group            0.1709   0.1512   -0.0197 ***   (-4.73)
Control sample        0.1653   0.1584   -0.0069       (-1.54)
(Main-control)                          -0.0128 ***   (-3.93)

*** Significant at the 0.01 level.