Office of Economic Analysis:
Report on the Comparison of Order Executions Across Equity Market Structures

January 8, 2001

Executive Summary

Background

The Report uses customer order data, which has only recently come available for Nasdaq securities. This enables accurate calculation of effective and realized spreads and order execution speeds, and also allows separate analysis of different types of orders. For the most part, the Report uses measures that will soon be available directly from market centers under Securities Exchange Act Rule 11Ac1-5 -- the Commission's recently adopted Execution Quality Disclosure Rule.

Obviously, any comparative analysis of market structures is necessarily complex. In spite of the fact that every attempt has been made to design the tests and present the results in a clear and unbiased manner, there are three important caveats that should be carefully considered by anyone reviewing this Report:

• There is no single, all-encompassing measure of execution quality. For example, although effective spread is an important component, some investors may prefer a fast execution at a guaranteed price (often available for small orders in Nasdaq securities at dealer market centers) to a slower execution with the possibility of price improvement (often available for small orders on the NYSE). In addition, effective spread measures the handling of a single trade, without considering the ability of a market structure to absorb a series of trades with minimal price volatility.

• Due to feasibility considerations, the size of the sample is somewhat limited. The one-week period of June 5, 2000 to June 9, 2000 covers a single, relatively tranquil episode that followed a period of higher stress and volatility. The results may differ under a different set of circumstances.

• Although the Report uses both matched-sample and regression techniques to try to control for the differences between the stocks that are listed on Nasdaq and the NYSE, these controls can never be perfect. Thus, there is always the possibility that the reported results are driven by remaining differences between the stocks rather than by differences in the degree of order interaction between the two market structures.

Methods

The results are separately calculated for four categories of Nasdaq stocks:

1)A group of the very largest Nasdaq stocks in terms of trading volume and market capitalization, as specifically selected by Nasdaq

2)A random sample of Nasdaq stocks with market capitalization over $1 billion

3)A random sample of Nasdaq stocks with market capitalization between $200 million and $1 billion

4)A random sample of Nasdaq stocks with market capitalization less than $200 million

For easy reference throughout the Report, these categories are referred to as "very large", "large," "middle," and "small," respectively.

In selecting the stocks for the Report, some initial filters were applied to ensure the availability of adequate historical information. The data in the report include customer orders for the week of June 5-9, 2000 taken from the Order Audit Trail System (OATS). These orders are for a total of 221 Nasdaq stocks, 25 of which were specifically selected by Nasdaq as being their top stocks in terms of trading volume and market capitalization (the very large category). The Commission staff selected the remaining 196 Nasdaq stocks by taking a random sample of Nasdaq stocks stratified by dollar trading volume. The Report includes data from the System Order Data (SOD) file for all 1141 NYSE stocks that pass the same initial filters used to construct the Nasdaq sample.

The Report includes a "matched pair" analysis that compares order executions in Nasdaq-listed stocks to NYSE order executions in NYSE-listed stocks, where the stocks in each pair have similar market capitalization, share price, return volatility and trading volume. For each measure, the first test uses only the 58 pairs that have the smallest aggregate differences across the four criteria. This analysis is complemented and confirmed by eleven other tests that use larger samples and use regression techniques to control for differences in these and other features. Seven of these tests rely entirely on regression techniques to make comparisons across the two samples, without any need to consider specific matched pairs.

The average market capitalizations for these 58 matched pairs of Nasdaq and NYSE-listed stocks are shown below, broken down by category.

Market Capitalizations for Matched Pairs

Results

The Report includes a variety of measures of order executions, including effective spread, realized spread, quoted spread, speed of execution, and rate of execution. This summary focuses on effective and realized spread for small market orders, and on speed of execution for market orders.

The effective spread measures the execution cost paid by investors by comparing the execution price to the midpoint of the NBBO quoted spread at the time that the order arrived at the market center for execution. These cost differences are doubled in order to make the effective spread statistics comparable to quoted spreads. For example, if a buy order arrives when the spread midpoint is $20 per share and the buyer pays $20.125 per share, the effective spread is (20.125 - 20) times 2, which equals $.25 per share.

The Report finds that, for market orders of 100-499 shares in the very large category of matched Nasdaq and NYSE stocks, the average effective spreads are nearly equal. (Although the first matched-pairs test for this category shows Nasdaq effective spreads lower than NYSE effective spreads by 1.2 cents per share, this estimate is statistically insignificant and the results are mixed across the range of the tests.) For 100-499 share market orders in the large, middle and small categories, the first matched-pairs test shows that the average Nasdaq effective spreads are from 5.7 to 11 cents per share wider than those for the matched NYSE stocks. (These differences are statistically significant and consistent across the range of tests).

The higher effective spreads for 100-499 share Nasdaq market orders, for all but the very large stocks, arguably might be explained by more difficult order flow. In other words, the dealers or other traders who are supplying liquidity on Nasdaq might be forced to charge wider effective spreads to protect themselves against a high proportion of informed trades included in the market orders. To test this possibility, we also examine realized spreads. The realized spread is similar to the effective spread, except that it uses the midpoint of the NBBO quoted spread five minutes after the order was executed. As its name implies, the realized spread is a very short-term proxy for the potential profit realized by the dealer or other trader taking the other side of the order. For market orders of 100-499 shares, the Report finds that average realized spreads are nearly equal for the very large matched Nasdaq and NYSE stocks. In the large, middle and small categories, the average Nasdaq realized spreads for 100-499 share market orders are 6.8 to 14.6 cents per share wider than those for the matched NYSE stocks. These results suggest that the higher effective spreads on Nasdaq are not a result of more difficult order flow.

The following table summarizes the results discussed above for 100-499 share market orders, using the first test based on the 58 closest matches. Tables 6 and 9 in the Report show that the other eleven tests produce substantially similar results.

Effective and Realized Spreads for 100-499 Share Market Orders

*,**, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively

To provide a comparison with the effective spreads paid by investors, the Report also examines quoted spreads. In the very large category, the Report finds that quoted spreads on Nasdaq stocks are on average 5.4 cents per share narrower than quoted spreads for the matched NYSE stocks. Comparing quoted spreads to effective spreads suggests that in Nasdaq stocks many orders are executed at the quotes, whereas many orders sent to the NYSE are executed at better prices inside the quotes. The NYSE rules require that the orders be given the opportunity to interact with other orders, which can result in price improvement. In the large and small categories, quoted spreads are nearly equal across the two markets, whereas the average Nasdaq quoted spreads are somewhat wider in the middle category. As is the case for the very large stocks, the Nasdaq quoted spreads for the other three categories are roughly equal to the effective spreads, whereas the effective spreads for small market orders sent to the NYSE reflect substantial price improvement.

NBBO Quoted Spreads

* and ** denote statistical significance at the 10% and 5% levels, respectively

The Report finds that market order executions are generally faster on Nasdaq than on the NYSE for 100-499 share orders. The difference disappears for the 500-1999 share market orders. Our results indicate that the NYSE executions tend to be somewhat faster than the Nasdaq executions for 2000-4999 share market orders, but Nasdaq believes that many large "not held" orders are not properly identified in their system. This miscoding may reduce the accuracy of the comparison between the two markets for the largest category of orders. The following Table shows the results for the first matched-pairs test. As shown on Tables 19-21 of the Report, the other tests yield similar results.

Execution Times in Seconds

*,**, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively

The Report

View the report (available in PDF format).