============================================================================
RUTGERS UNIVERSITY
FACULTY OF MANAGEMENT
Department of Finance and Economics
180 University Avenue
Newark, NJ 07102-1897
David K. Whitcomb
Professor of Finance and Economics
O: (973) 353-5264 R: (212) 228-2388
e-mail: whitcomb@andromeda.rutgers.edu
April 13, 1998
Mr. Jonathan Katz
Secretary
Securities and Exchange Commission
450 Fifth Street, NW
Washington, DC 20549
Re: File No. SR-NASD-98-21
Dear Secretary Katz:
I am Professor of Finance in the Graduate School of Management of Rutgers
University, one of the earliest academic researchers in the area of
securities market microstructure, and co-author of The Microstructure of
Securities Markets (Prentice-Hall, 1986) and numerous other journal articles
and monographs on market microstructure. I testified before the House
Commerce Committee regarding decimalization, and my comments regarding NASD
governance, various versions of the NAQCESS proposal, the order handling
rules and other NASDAQ issues have been reported by CNN, NPR, PBS, AP,
Business Week, The New York Times, Securities Week, The Washington Post, The
Wall Street Journal, and other publications and cited in previous SEC
Decisions. I am also founder & CEO of Automated Trading Desk, Inc., which
developed the first expert system for automated limit order trading. In
1997, ATD systems traded over $20 billion of Nasdaq and NYSE stocks for
clients ranging from a large investment bank to small brokerage firms.
Accordingly, I have academic, practical and personal knowledge of and
interest in the issue presently before the Commission; indeed, I wrote 4
comment letters on this issue in 19971 , which I incorporate by reference.
EXECUTIVE SUMMARY: The NASD seeks to eliminate mandatory minimum market
maker proprietary quotation size requirements ("mandatory quote sizes") on
all Nasdaq stocks, a proposal they call the Actual Size Rule ("ASR"). Their
current rationale for the ASR proposal is the economic theory argument that
the ASR, by reducing market maker quotation risk, may increase market maker
participation and reduce bid-asked spreads. To this theoretical argument are
added two NASD empirical studies2 of the so-called Pilot Test of the ASR3,
which purport to show that the ASR does not increase spreads or volatility
or reduce liquidity or access to the SOES system (a trading system essential
for immediate market access by customers whose brokers will not or cannot
sell their orders directly to dealers). There are 3 crucial flaws in the
NASD's rationale for the ASR:
1 These letters are Whitcomb to Lindsey re SR-NASD-96-43, January 7,
1997 ("Whitcomb January 7, 1997 letter"); Whitcomb to Katz re SR-NASD-97-26,
May 12, 1997 ("Whitcomb May 12, 1997 letter"); Whitcomb to Katz re
SR-NASD-97-26, July 3, 1997 ("Whitcomb July 3, 1997 letter"); and Whitcomb
to Katz re SR-NASD-97-49, August 12, 1997 ("Whitcomb August 12, 1997 letter").
2 "Effects of the removal of minimum sizes for proprietary quotes in
the Nasdaq Stock Market, Inc.," prepared by the Staff of NASD Economic
Research, June 3, 1997 ("First NASD Study") and "The effect of the Actual
Size Rule on the Nasdaq Stock Market, Inc.," prepared by the Staff of NASD
Economic Research, March 4, 1998 ("Second NASD Study").
3 This test was conducted in two phases. Phase 1 subjected 50
stocks, covering over 35% of Nasdaq dollar volume to the ASR, while in Phase
2 the pilot group was expanded to 150 stocks by adding stocks randomly
selected from NMM stocks with SOES tier sizes of 1000 and 500 shares
(excluding all stocks in the First 50 and Second 50). The analysis in both
phases is marred by the fact that the 10 largest Nasdaq stocks have all been
subject to the ASR since January 20, 1997, and the next 10 stocks have not,
making it impossible rigorously to determine the effect of the ASR on the
large stocks which account for a very significant portion of volume,
especially at the retail level. Despite my pleas to the Commission (see
Whitcomb May 12, 1997 letter and Whitcomb August 12, 1997 letter), this
fundamental flaw in the test design has never been corrected.
•NASD bases its economic theory rationale for ASR on the assumption that
Nasdaq market makers engage in perfect quote price competition. This
assumption is completely inappropriate, as is shown by the pioneering
Christie & Schultz paper4, by the Commission's Section 21 (b) Report and by
a study of quotation price behavior in September and October 1997 which I
co-authored5. Under the more realistic alternative economic model of
oligopolistic quote price behavior by market makers, there is no reason to
expect that market makers will participate in greater numbers or reduce the
equilibrium market spread in response to a risk-reducing policy change like ASR.
•NASD's own empirical studies produce data inconsistent with their economic
model's predictions. Both the First and Second NASD Studies show that market
maker participation rates and bid-asked spreads are unaffected by existence
of the ASR. These results are consistent with the results predicted under
the alternative model of oligopolistic quote price behavior by market
makers. They suggest that the only clear beneficiaries of the ASR are the
market makers, not individual or institutional investors.
•NASD asserts that the ASR does not reduce the Nasdaq market's depth or
liquidity and does not reduce access to automated execution of market orders
via the SOES system (mandated by the SEC following the October 1987 crash).
However, their own data and the results of numerous studies by others6 show
that the ASR does indeed reduce market depth, liquidity and SOES access.
Both the Second NASD Study and the Simaan-Whitcomb Study show that the
reduction in market depth and liquidity are particularly pronounced at times
of market stress, such as the "market break" of October 27-28, 19977
4 Christie, W.G. and P.H. Schultz, 1994, "Why Do NASDAQ Market
Makers Avoid Odd-eighth Quotes?" Journal of Finance, 49, 1813-1840
("Christie-Schultz Study").
5 Y. Simaan & D. Whitcomb, "The Quotation Behavior of ECNs and
Nasdaq Market Makers," NJCRFS Working Paper 98-01, December 1997
("Simaan-Whitcomb Study", attached as Appendix 1).
6 See Barclay, Christie, Harris, Kandell and Schultz, "The Costs of
Trading Nasdaq Issues: The Impact of Limit Orders and ECN Quotes," Working
Paper 97-10, February 1998 ("Barclay, et. al. Study"); ETA and
DRI/McGraw-Hill, "Analysis of Automatic Executions on the Nasdaq Stock
Market," May 1997 ("DRI/McGraw-Hill Study", attached to Comment Letter from
Richard Y. Roberts re SR-NASD-97-26, dated May 12, 1997); and Simaan -
Whitcomb Study.
7 The Second NASD Study shows that market maker average aggregate
quotation size is always smaller for First 36 (ASR) stocks than for Second
36 (non-ASR) stocks (about an 88% ratio), but it fell nearly 9% during the
two days October 27-28 vs. a negligible decline for Second 36 stocks. As I
show in the detailed analysis in Section II below, these data conceal the
real problem by inappropriate aggregation. The Simaan-Whitcomb Study (Table
3, Panel B) shows that average aggregate market maker bid size for ASR
stocks fell to only 76% of that for non-ASR stocks on October 27 (the day
the market fell 560 points), and the ratio of ASR/non-ASR average depth was
at its lowest during the half-hour periods of maximum downward stress.
Market makers still enjoy significant privileges and subsidies in the Nasdaq
marketplace. These include:
•Minimum 'tick' sizes. ECN and market maker quotes priced at intervals
smaller than 1/16th are not permitted in the Nasdaq quote montage. This is a
subsidy to dealers of huge magnitude (estimated at billions of dollars/annum
by former Commissioner Wallman in his testimony to Congress on
decimalization), as it inhibits price improvement on hundreds of thousands
of retail market orders filled each day at the NBBO.
•Free advertisement of dealer 'indications of interest' via the Nasdaq quote
montage, paid for by NASD Member dues and Nasdaq transaction fees charged
order entry firms.
•Reduced margin requirements available to Nasdaq market makers.
•Exemption from short sale limitations.8
•New privileges proposed in the NODES system including: a 17-second 'backing
away window' during which market maker quotes are visible but not hittable,
a 10-second-hold period during which limit orders (but not market maker
quotes) may not be cancelled, sole right to sponsor institutional access to
NODES.
Once these special privileges and subsidies have been eliminated, it would
make sense also to eliminate any market maker responsibility to post quotes
of any minimum size. However, until this happens, eliminating that
responsibility will just reward market makers without any benefit to the
trading public.
Although the ASR would be simply a "gift" to market makers without provable
public benefit, it would create specific harm:
•ASR would reduce automated access to the market, especially at times of
market stress (as it did during October 27-28 1997).
•ASR would increase the pressure on brokerage firms to sell retail order
flow to dealers as the only practical way to guarantee their customers
access to automated executions, especially at times of market stress. The
percentage of retail order flow delivered to dealers pursuant to payment for
order flow or internalization arrangements is already very high and will be
increased in the long run by restricting access9.
•An increase in preferenced retail order flow will, in turn, reduce quote
price competition.
•ASR will make the "signals" market makers give to the market (one of their
most valuable economic functions) far "noisier" by reducing the cost of
erroneous or misleading signals.
8 Only Œprimary market makers¹ are exempt from the Œbid tick test¹
for short sales. However, all Nasdaq market makers are classified as primary
market makers.
9 Table III.11 of the Second NASD Study reports that 35 40% of all
Nasdaq trades and around 30% of share volume are generated by "proprietary
autoexecution systems". The report makes much of the fact that there was no
difference in the autoexecution statistics resulting from imposing the ASR
when the 18 days before November 10, 1997 were compared to the 20 days
following. That is completely un-surprising, since implementing purchase of
order flow or internalization arrangements requires contractual and computer
system arrangements which take months to effect.
The Commission acted decisively and effectively to force increased
competition on the Nasdaq market with the Order Handling Rules and is now
searching for ways to make the listed market more competitive (via changes
discussed in the "Concept Release"). It would be a shame if the SEC were to
approve a step that will reduce competition.
• Analysis of NASD Theoretical Arguments for ASR
The NASD has based its theoretical argument for the ASR on 3 propositions:
The First NASD Proposition is that, in an order-driven market (which they
initially assumed would occur on January 20, 1997, but now merely predict
for the future), the regulatory reason for requiring market makers to
provide a minimum of liquidity will disappear. To see how absurd this is,
assume, arguendo, that the Nasdaq market does become entirely order-driven
tomorrow. In that event, the alleged "need" for maintaining one-sixteenth
(or any) minimum price increments — subsidizing market makers in order to
retain the liquidity they provide — would immediately disappear. In other
words, if the market becomes order-driven and there is no need to maintain
market maker liquidity, then the minimum price increment should drop to one
penny the same day. I have seen no proposal from NASD to move immediately to
decimal pricing with penny increments. Indeed, in its NODES filing, the NASD
speaks of "maintaining incentives for market makers"10 and actually proposes
several new "incentives" in that filing. Talk about a diminished regulatory
reason for requiring market makers to provide minimum liquidity is just that
— talk.
The Second NASD Proposition is that "Stock Exchange specialists have 100
share minimums, so why shouldn't Nasdaq market makers?" This argument
ignores a fundamental difference between the two markets. The primary
markets for listed shares, the NYSE and the AMEX, have always given limit
orders priority over specialist quotes at the same price. This means that a
limit order placed in either market has a reasonable chance of executing. By
contrast, in the Nasdaq market, even under the Order Handling Rules and the
NODES proposal, market makers receiving preferenced market orders (or a
phone call from an institutional broker) will be able to trade ahead of all
limit orders at the inside (except those given to the particular market
maker). For example, if there are 30 dealers in a particular stock, at least
29 of them will be able to trade ahead of my limit order (and all 30 if my
order is posted on NODES or an ECN). This is part of the fact that Nasdaq
dealers have many privileges (minimum increments, free advertising, lower
margin requirements, relief from the short sale rule, and no "affirmative
obligation" to stabilize), and should, in return, have a responsibility to
make firm, meaningful quotes.
The Third NASD Proposition is the one on which they currently base their
main argument for the ASR. In the NASD's words, "economic theory indicates
that permitting dealers to quote in size commensurate with their true
trading interest could further narrow quoted spreads and enhance the pricing
efficiency of the Nasdaq marketplace."11 The NASD does not spell it out, but
the particular economic theory on which they base this proposition is the
theory of perfect competition. It is true that, in perfectly competitive
long run equilibrium, if market makers are held to firm, meaningful quotes
in the post-Order Handling Rules environment, their profits would be
diminished enough to reduce the amount of liquidity they are able to supply.
The issue here is whether market makers need to be subsidized by eliminating
their firm quote obligation (or by never having to face decimal pricing) in
order to keep them in business. The answer is "absolutely not", for several
reasons:
10 See Federal Register, v. 63, no. 48, p. 12124.
11 See Federal Register v. 63, no. 55, p. 13896.
First, most market makers, especially the top ten or so "lead market makers"
who supply most of the Nasdaq market's liquidity through their institutional
brokerage business, are enormously profitable. To put it technically, market
makers as a group are earning oligopolistic profits, in part because of the
practices described in the SEC's Section 21 (b) report (which my research
with Simaan show to be continuing12) and in part because of their ability to
purchase or internalize retail order flow13. Anything that reduces their
profits, whether it be the Order Handling Rules, the Firm Quote obligation,
decimal pricing, or the activities of day traders, merely reduces the
subsidy they are enjoying as a result of the NASD's past successes as a
trade organization, operating under New Deal structures, in protecting this
oligopoly. Only when an industry is actually in perfectly competitive
equilibrium does an action which reduces profits necessarily drive some
firms out of business and reduce the quantity of "goods" produced; market
making cannot be so described.
Second, if market makers wish to reduce the risk of unwanted trades in an
environment of narrower spreads and a continued Firm Quote obligation, they
need only make more careful markets. Any market maker who fears being "hit"
by a 1000 share order need only move his quote 1/16th or 1/8th away from the
"inside" to eliminate the risk.
Third, by enabling a market maker to put customer limit orders ahead of
their quotes, the SEC has taken an action that reduces market maker risk. In
other words, we have a self-equilibrating system: As the use of limit orders
increases, narrowing spreads, dealer risk is replaced by limit order trader
risk; and the dealer's risk-reward tradeoff is left essentially unchanged.
Dealers may be required to supply less of the liquidity, but they will take
less risk.
Last, one need only look at the listed stock market to see that market
makers will not be driven out of business by narrow spreads and limit order
exposure. In both the NYSE and AMEX markets, stocks have for a long time had
substantially narrower average spreads than comparable Nasdaq stocks. In
these markets, limit orders have always been exposed (and even have priority
over specialist own-account trades) and market orders are exposed to gain
price improvement whenever the spread is more than the minimum tick size.
Despite these features, specialists, "upstairs" market makers and third
market dealers all are able to stay in business and thrive. There is clearly
no need to worry about the economic viability of market makers when some
elements of fair competition are introduced to the Nasdaq market.
12 See Y. Simaan & D. Whitcomb, "The Quotation Behavior of ECNs and
Nasdaq Market Makers," NJCRFS Working Paper 98-01, December 1997 (attached
as Appendix 1). In this paper, we show that, while the Order Handling Rules
have narrowed spreads, they have not caused market makers to abandon their
practice of avoiding "odd increments". We find that leading market makers
only quoted odd sixteenths 6.9% of the time (vs. an expectation of 50%).
Even when a market maker is alone at the inside odd sixteenths are quoted
only 11.8% of the time (vs. 49.7% of the time for ECNs). Only in an
oligopolistic industry could this practice be maintained.
13 NASD commissioned some economic studies in conjunction with their
earlier attempt to refute the Christie-Schultz Study and promote NAQCESS and
its predecessors which presented Herfindal Indexes, Concentration Ratios,
and other measures purporting to show that market making fits the definition
of a perfectly competitive industry (basically because there are many firms
in the industry). This is a gross misuse of elementary economics because it
entirely ignores the mechanism by which price maintenance is achieved in an
industry with many rival firms. The work of Christie and Schultz, the SEC¹s
21 (b) report on the "culture of collaboration", and the Simaan-Whitcomb
Study¹s evidence of continued market maker avoidance of "odd increments"
show that the industry is more accurately characterized as oligopolistic
price leadership.
• Assessment of Empirical Evidence of the Effects of the ASR
The essential point to make about the empirical evidence is that none of it
is consistent with the predictions of the NASD's "economic theory" rationale
for the ASR. The reason is that the NASD's particular economic theory
doesn't fit the Nasdaq market. On the other hand, the empirical evidence is
consistent with what one would expect under the alternative economic model
that the Nasdaq market remains an oligopoly where market makers only rarely
engage in genuine quote price competition.
I will not waste the Commission's time going over the evidence relating to
the NASD's "economic theory" prediction that the ASR will lead to increased
market maker participation and reduced spreads. Everyone, including the
authors of the First and Second NASD Studies, agrees that, 1 _ years after
the Order Handling Rules were implemented, the ASR has neither increased
market maker participation nor reduced spreads relative to stocks not
subject to the ASR. Market maker participation has increased slightly over
time and just as much for non-ASR stocks as for ASR stocks, refuting the
doomsday assertion that market makers would leave the business in droves if
they were actually forced to make markets in the "New World Order" brought
about by the Order Handling Rules. Spreads fell literally overnight when
each "cohort" of stocks was brought under the Order Handling Rules and again
when the minimum increment was reduced from 1/8th to 1/16th, but this had
everything to do with the new quote price competition resulting from the ECN
Display Rule and nothing whatsoever to do with the ASR or the willingness of
market makers to engage in quote price competition.
The most interesting and controversial empirical evidence relates to the
effect of the ASR on market depth, volatility, and on access to the SOES
system (the main means by which individual investors whose brokers do not
sell or internalize their order flow obtain immediate executions in the
Nasdaq market). Here, both the First and Second NASD Studies assert that ASR
has done no harm; whereas other researchers and the NASD's own data show
that imposition of the ASR reduces market depth, increases market
volatility, and reduces access to SOES.
How do the First and Second NASD Studies manage to turn a sow's ear into a
silk purse? By the art of regression analysis. Both studies use essentially
the same multivariate regression model14 which purportedly seeks to control
for the effects of changes in trading volume and volatility when changes in
measures such as spreads, depth (aggregate market maker quoted size at the
inside), or SOES volume are related to the presence or absence of the ASR.
However, this model was seriously flawed when it was used in the First NASD
Study, and it remains seriously flawed, as the results of as a report by
financial econometrician Professor Yusif Simaan of Fordham University
demonstrate15. The fundamental flaw in the NASD's regression model is the
fact that the contemporaneous spread, depth, or SOES volume may be both
determined by and determining of trading volume and volatility. Hence the
use of contemporaneous volume or volatility as a " control" variable may
tend spuriously to reduce the effect of the dummy variable (presence or
absence of the ASR) on the measure of change in spread, depth or SOES
volume. Professor Simaan suggested a simple feasible correction in the
flawed methodology, but NASD Economic Research never implemented it. As a
result, their regression analysis has no probative value. In the words of
former Commissioner Roberts, "the multivariate regression analysis…in the
NASD Study appears to be designed more to 'control' results than to
'control' for exogenous factors."16
14 See, for example, Second NASD Study, Appendix A, pp. 100-101.
15 See "The econometric methodology in (First NASD Study) ŒAppendix
B: Formal Statistical Analysis¹" by Yusif Simaan, July 1997. This was
appended as Attachment I of Richard Y. Roberts comment letter on
SR-NASD-97-26, dated July 2, 1997, but since the NASD ignored it in
preparing their Second NASD Study, I attach it again here (as Appendix 2).
16 See Roberts, Op. Cit., p. 3.
The direct evidence (not processed through a flawed regression model) shows
clearly that several key measures of Nasdaq market quality have been
deleteriously affected by the ASR. These measures and the main results include:
Market depth in normal times. In most studies, depth is proxied by the
(time-weighted) average aggregate market maker quotation size at the inside.
The main results are:
•The First NASD Study shows that average aggregate market maker quoted size
at the inside has declined 27.5% for First 40 stocks;17 adding ECN quoted
size at the inside leaves First 40 (ASR) stocks with a 17% decline in
liquidity available through SOES and SelectNet preferencing vs. only a 4%
decline for Second 40 (non-ASR) stocks.
•The Simaan-Whitcomb Study shows that, as of September 1997, average First
40 aggregate bid sizes were only 85% of Second 40 size; for ask quotes, the
figure was 80%18.
•The Second NASD Study shows that, just prior to the October 27, 1997 market
break, First 36 (ASR) stocks had average aggregate dealer quote sizes 88% as
large as Second 36 (non-ASR) stocks.
Market depth during market stress. A major fear of those opposing the ASR
has been that, when the chips are down, market makers will take advantage of
the ASR to avoid making markets in meaningful size on the side of the market
where the stress appears. The October 27, 1998 market break gave us an
opportunity to test this proposition. The key findings are:
•The Second NASD Study shows that, whereas "normal" average aggregate dealer
quote sizes are 88% as large for First 36 (ASR) as for Second 36 (non-ASR)
stocks, they were only 81% as large during the two days of the crisis,
October 27 and 28, 1998. In other words, the NASD's depth measure fell 8.6%
for ASR stocks vs. only 0.4% for non-ASR stocks.
•The Simaan-Whitcomb Study shows that, whereas September aggregate average
bid sizes were 85% as large for First 40 (ASR) as for Second 40 (non-ASR)
stocks, on October 27, 1997, the relative percentage fell to only 76%19.
Disaggregating by half-hour period, we found that the smallest ASR/non-ASR
depth ratios occurred during the times of greatest downward pressure,
9:30-10:00, 10:-10:30, 11:00-11:30, and especially 3:00-3:30 PM, when ASR
stocks had only 67% of the depth of non-ASR stocks.
While both the Second NASD Study and the Simaan-Whitcomb Study show a clear,
economically significant, reduction in market depth for ASR stocks, the NASD
data do not look so bad. Why? Because the NASD data aggregate over side and
day. These data do not confront the real problem, namely that at the times
of greatest market stress, market makers will refuse to make markets in
depth on the side that is subject to the most stress. By aggregating bid and
ask sizes and including October 28 (when the market's general direction was
up), the NASD data tend to conceal, rather than to illuminate the problem.
17 The actual decline is even larger since the average includes
quotes of more than 1000 shares; frequently these quotes prove ephemeral
once the first 1000 shares have been taken.
18 Simaan-Whitcomb Study, Table 2, Panel A; the percentages are
similar but the aggregate sizes are smaller for the "truncated distribution"
data of Panel B.
19 Simaan-Whitcomb Study, Table 3, Panel B
Market volatility. There is no universally-agreed measure of volatility, but
a sensible ad hoc measure that NASD Economic Research uses is "normalized
effective depth", a trade-based liquidity measure akin to the economist's
"arc elasticity". Specifically "normalized effective depth" is the dollar
volume required to move the bid-ask midpoint 1% without a reversal of
direction; a high normalized effective depth means higher liquidity and
lower volatility.
•The First NASD Study shows that the ability of First 40 stocks to resist a
1% price move decreased by 6.4%, while the ability of Second 40 stocks to
resist a similar move actually increased by 16.9%.
•The Second NASD Study reports the same statistic for "Next 103" (ASR stocks
not in the First 50) vs. the non-ASR 3207. While normalized effective depth
is much smaller at every threshold for the 103 stocks even before they
became subject to the ASR, calling into question the sample selection
procedure, it declines substantially relative to the non-ASR 3207 at every
threshold once the 103 stocks become subject to the ASR20 .
SOES access. The First NASD Study provides no meaningful data on SOES
access21 and asserts (as does the Second) that the ASR does not diminish
access to or usage of SOES, but other studies show that, as would be
expected, the ASR does indeed diminish usage of SOES:
•Data from the Barclay, et. al. Study show that average SOES trade size for
the First 50 stocks has fallen from 930 before January 20 to 776 (after
February 9), while SOES trade size for Second 50 stocks has fallen only from
909 to 854.22 Most strikingly, the Barclay, et. al. data show that SOES
trades as a proportion of all Nasdaq trades fell from 24.9% to 19.5% for the
First 50, whereas they rose from 19.8% to 21.1% for the Second 50.
•A study I commissioned and supervised at the company I founded looked at a
major client firm's SOES order completion rates (SOES volume divided by
number of shares the firm entered into SOES) for the 6 weeks before January
20, 1997 versus the two months March and April 1997 after 150 stocks had
become subject to the Order Handling Rules. We found that the SOES order
completion rates for the 50 stocks subject to the Actual Size Rule fell from
54.4% before January 20 to 47.5% in March and April, whereas SOES order
completion rates actually rose from 59.6% to 64.1% for the "Second 50"
stocks still enjoying 1000 share dealer quote minimums.22
•The Second NASD Study did look at SOES volume and the results are stunning:
(1) In a year of huge increases in Nasdaq dollar volume, for First 36 (ASR)
stocks, SOES dollar volume has risen only 16.7%, but Second 36 (non-ASR)
SOES dollar volume has increased 94%. (2) First 36 SOES volume as a
percentage of all dollar volume has fallen from 13.4% to 9.8%, a decrease of
37%. (3) SOES average trade size has fallen 6.7% for First 36 (ASR) stocks,
while it has risen 2% for non-ASR stocks24 . (4) For the 103 stocks newly
added to the ASR set, the results are not so dramatic, but the Next 103
(ASR) stocks' average dollar SOES volume/stock was only $138,800 vs.
$201,700 on average for the 3207 stocks not subject to the ASR, and SOES
volume as a percentage of total dollar volume fell from 9.0% to 6.6% in the
20 days following implementation of the ASR25.
20 See Table III.8. Through the magic of multiple regression, NASD
Economic Research makes this difference disappear, but since their
regression model confuses logical causality, we cannot rely on it to explain
away the results on normalized effective depth any more than we can for
other market quality measures.
21 The First (and Second) NASD Study presents several largely
irrelevant statistics to support their claim that "implementation of the
Actual Size Rule has not diminished the ability of small investors to
receive automated executions through SOESŠ." Their statement that "99% of
SOES trades were executed at one price" is both gramatically and
economically confusing. This means that few individual SOES orders are
filled at two or more different prices. However, it is a long way from this
to the NASD¹s conclusion that "investors received SOES executions at the
size and price they desired." If a person enters a SOES buy order when the
offer is 22 _, but gets an execution 30 seconds later at 22 1/2 (after a 25
cent "chase"), the NASD¹s statistic will misleadingly imply that she "got
the price she wanted". The DRI/McGraw-Hill Study shows that such frustrating
chases are very common. Please see former Commissioner Roberts¹ comment
letter on SR-NASD-97-26 dated May 12, 1997 for detail. (The DRI/McGraw-Hill
results actually understate the extent to which ordinary investors have to
"chase" prices, since SOES day traders are more likely than other traders
just to cancel their orders when the quote changes.)
22 See Barclay et. al. Study, Table VI. Panels A and B.
23 The higher beginning order completion rate for the Second 50
stocks may be due to their higher average quoted size; in any event, the key
statistic is the relative change between the two groups. The "Third 50"
(stocks phased in on February 24) and a sample of non-phased-in stocks both
have order completion rates and before vs. after changes similar to the
Second 50. The data consist of all 174,221 SOES orders entered by the client
firm during the periods in question and were "captured" by computer systems
maintained by Automated Trading Desk, Inc. The test described here is
similar to the test DRI/McGraw-Hill ran on ETA-supplied data (please see
former Commissioner Roberts¹ comment letter dated May 12, 1997) except that
the DRI/McGraw-Hill measure was the percentage of fully executed orders.
They found the same result: the frequency of execution fell from early
January to mid-April for the First 50 stocks, whereas it rose for all other
stock groups.
24 Results (1) to (3) above are from Second NASD Study, Table B.2.
25 See Second NASD Study, Table C.13.
Neither appropriate economic theory nor the empirical evidence supports the
NASD's proposal to expand the Actual Size Rule to cover all Nasdaq stocks.
Market makers are fat and happy despite the reduction in their unit profits
following implementation of the Order Handling Rules. This is no time to
reduce market liquidity, hurt the ability of public investors to obtain
immediate execution, and drive retail brokers into the arms of purchasers of
order flow.
Sincerely yours,
David K. Whitcomb
Professor of Finance, Rutgers University
President/CEO, Automated Trading Desk, Inc.
Attachments:
Appendix 1 Simaan-Whitcomb Study
Appendix 2 Simaan Analysis of NASD Econometric Methodology
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
James H. Lee (713)706-3300 x227 (713)977-7975 FAX jlee@soes.com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
President, Momentum Securities Management Company, Houston, TX
http://www.soes.com
President, Electronic Traders Association, Washington, DC
http://www.electronic-traders.org
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~