The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
DID THE GOOD GUYS LOSE? HETEROGENEOUS TRADERS
AND REGULATORY RESTRICTIONS ON DUAL TRADING
by
Peter R. Locke, Asani Sarkar and Lifan Wu
Federal Reserve Bank of New York
Research Paper No. 9611
May 1996
This paper is being circulated for purposes of discussion and comment only.
The contents should be regarded as preliminary and not for citation or quotation without
permission of the author. The views expressed are those of the author and do not necessarily
reflect those of the Federal Reserve Bank of New York or the Federal Reserve System.
Single copies are available on request to:
Public Information Department
Federal Reserve Bank of New York
New York, NY 10045
Did the Good Guys Lose?
Heterogeneous Traders and Regulatory Restrictions on Dual Trading
Peter R. Locke
Commodity Futures Trading Commission
Asani Sarkar
Colwnbia University
and
Federal Reserve Bank of New York
Lifan Wu
City University, Hong Kong
Tel: 202-418-5287 (Locke)
212-720-8943 (Sarkar)
Fax: 202-418-5527 (Locke)
212-720-1773 (Sarkar)
Email: ASARKAR@PIPELrNE.COM
Previous version: March, 1996
This version: May, 1996
Sarkar and Wu are grateful to the Office for Futures and Options Research of the University
of Illinois at Urbana-Champaign for financial support, and to the Commodity Futures Trading
Commission and the Chicago Mercantile Exchange for the provision of data. The views
stated herein are those of the authors and do not necessarily reflect the views of the Federal
Reserve Bank of New York, the Federal Reserve System or the Commodity Futures Trading
Commission or their respective staffs. All errors and omissions are our responsibility alone.
ABSTRACT
Did the Good Guys Lose?
Peter R. Locke
Asani Sarkar
Lifan Wu
JEL Classification number: 612
We study the effect of restrictions on dual trading in futures contracts. Previous studies
have found that dual trading restrictions can have a positive, negative, or neutral effect on
market liquidity. In this paper, we propose that trader heterogeneity may explain these
conflicting empirical results. We find that, for contracts affected by restrictions, the change in
market activity following restrictions differs between contracts. More important, the effect of
a restriction varies among dual traders in the same market. For example, dual traders who
ceased trading the S&P 500 index futures following restrictions had the highest personal
trading skills prior to restrictions. However, realized bid-ask spreads for customers did not
increase following restrictions. Our results imply that securities regulation may adversely
affect customers, but in ways not captured by broad-based liquidity measures, such as the bidask spread.
1. Introduction and Background
Dual trading is the practice whereby floor traders on a futures exchange execute trades for
their proprietary accounts and customers on the same day. There is considerable confusion in
academic as well as policy circles regarding the desirability of dual trading. Proponents of
dual trading believe it enhances market liquidity; opponents emphasize the possibility of
trading abuse. The empirical evidence is equally confusing. Depending on the market
studied, the correlation between dual trading and liquidity may be negative, positive or zero
(Fishman and Longstaff 1992, Smith and Whaley 1994, Chang and Locke 1996.) It is,
perhaps, natural that, given the lack of clear results, policy makers have chosen to emphasize
the potential for trading abuse when dual trading occurs. In 1992, the U.S. Congress passed
the Futures Trading Practice Act, which, among other things, compelled the CFTC to pass
regulations to prohibit dual trading on high volume contracts. 1
In this research, we seek to explain the conflicting empirical results surrounding dual
trading. We propose that floor traders in general, and dual traders in particular, are
heterogenous with respect to trading and execution skills. Our approach is motivated by two
strands of the existing theoretical literature on dual trading. 2 In one strand, dual trads:rs are
'
modeled as skilled brokers and market makers, whose existence contributes a potentially
positive effect on liquidity (Grossman, 1989). In the other strand of literature, dual traders
are viewed as informationally-motivated traders, with a potentially negative effect on liquidity
(Fishman and Longstaff, 1992; Roell, 1990; Sarkar, 1995). 3 In markets with a relatively high
proportion of skilled dual traders, a restriction on dual trading may harm liquidity. In markets
with a relatively low proportion of skilled dual traders and a relatively high proportion of
1
informationally-motivated dual traders, there may be a zero or negative
correlation between
dual trading and market liquidity.' It is likely that each futures pit contai
ns a different mix
of dual traders, and is affected differently by the same dual trading restric
tion. Thus, trader
heterogeneity could potentially explain the conflicting empirical results
surrounding dual
trading.
We examine two particular episodes of dual trading restrictions: the Chicag
o Mercantile
Exchange's (CME) top-step rule', which restricted dual trading in the
S&P index futures after
June 21, 1987; and rule 5526, which the CME imposed on dual traders
in high volume
contracts in 1990. To study the effects of rule 552, we examine the
Japanese Yen futures
contract. 7 The wording of the two rules (top-step and 552) appears differe
nt, but their effect
on dual trading is similar. The top-step rule specifically bans dual trading
only on the topstep of the pit, whereas rule 552 bans dual trading in all active contra
cts. However, the
geography of the pit dictates that brokers stand on the top step of the
pit to maintain sight
contact with their clerks and the trading desk. By banning personal trading
on the top step,
the top-step rule severely constrains brokers from dual trading.
The main steps of our analysis are as follows. First, we document the
occupational choice
of dual traders following restrictions (eg., whether a dual trader becam
e a pure broker
following restrictions). Next, we group dual traders prior to restrictions
according to their
observed occupational choice following restrictions (e.g., one group consis
ts of those dual
traders in the pre-restriction period who chose to become pure broker
s following restrictions).
We establish heterogeneity between these groups of dual traders with
respect to their prerestriction trading patters, as well as their trading and executions skills.
Finally, we examine
2
whether customers in the aggregate are hurt by dual trading restrictions.
Our empirical results document the existence of heterogeneity among dual traders, and
demonstrate that dual trading restrictions affect different groups of dual traders in the same
market differently. For example, the group of dual traders who chose to become pure brokers
(locals8) following restrictions was primarily engaged in trading for customers (their own
accounts) prior to restrictions. In addition, for the S&P 500 futures, dual traders with the
highest personal trading skills quit trading this contract following restrictions. However, the
exit of highly skilled traders did not increase customers' trading costs following restrictions.
Two possible reasons are: one, execution skills required for personal trading are not highly
correlated with execution skills required for customer orders. Two, we find evidence that, for
the S&P 500, dual traders with the highest customer order execution skills chose to become
pure brokers following restrictions.
For regulatory policy, an implication of our results is that restrictions on dual trading are
difficult to justify on the basis of economic arguments. The effect of restrictions on the
market is at best neutral, with side effects that are potentially negative for customers. Broadbased measures of liquidity (such as the bid-ask spread), used in previous studies, fail to
capture these side effects of regulation. For the contracts we study, this harmful side effect
may arise from customers being deprived of the services of the most skilled traders. In our
study, this has no apparent adverse effect on customers' trading costs--but, for other contracts,
a different outcome could occur. Perhaps, concerns about trading abuses are best countered
by increasing competition among brokers9, and a better technology for recording trades, not
by restricting dual trading.
3
As mentioned earlier, the empirical evidence on the efficacy of dual trading restrictions is
mixed and confusing. The Commodity Futures Trading Commission (CFTC) (1989) finds no
evidence of superior trading skills for dual traders, in a variety of markets. Dual traders
appear to supply liquidity with their personal trading, but their personal trading is no more
important than traders who are exclusive proprietary traders. Fishman and Longstaff (1992)
find some evidence that customers of dual traders have higher trading profits (lower trading
costs) compared to customers of pure brokers. Chang and Locke (I 996) find that dual traders
are specialized, concentrating either on brokering or personal trading. Dual traders do not
earn as much with their personal trading as do exclusive personal traders. They have mixed
results on whether their customers have lower trading costs compared to customers of pure
brokers.
In addition to skill levels, the literature discusses the effect of dual trading on liquidity.
Smith and Whaley (1994) and Walsh and Dinehart (1991) find some evidence consistent with
the notion that the practice of dual trading increases liquidity. Smith and Whaley (I 994) base
their findings on a particular bid ask spread estimator, which, Locke and Venkatesh (1996)
show bears little relation to actual customer transactions costs. Chang and Locke ( 1996) find
that a restriction on dual trading has a positive effect on liquidity.
The existing theoretical literature on dual trading focuses on two different aspects of dual
trading. Grossman (1989) asserts that dual traders are superior at order execution and market
making. Also, dual traders provide flexibility by reacting quickly to changing market
conditions, absorbing excess order flow, or racing to fill customer orders. By competing with
both pure brokers and market makers (at least in the short run), dual traders enhance market
4
liquidity. Fishman and Longstaff (I 992), on the other hand, model dual
traders as mimicking
the trading decisions of informationally-motivated traders and reducing
trading profits of
informed traders. Roell (1990) and Sarkar (I 995) show that dual traders
may also hurt
uninformed traders and, as a consequence, liquidity may decrease.
The rest of the paper is organized as follows. Section two describes the
data and our
procedure for identifying various floor trader groups. In section three,
we describe the effect
of restrictions on market activity. Section four documents the occupa
tional choice of dual
traders following restrictions. In section five, we estimate relative execut
ion skills of dual
traders prior to restrictions and the effect of restrictions on customer trading
costs. Section
six concludes.
2. Data and Sample Descriptions
The Computerized Trade Reconstruction (CTR) data is used in this study
for two futures
contracts which trade on the CME, the S&P 500 index and the Japane
se Yen. The data were
generously supplied by the CFTC, and consist of detailed records for every
transaction on the
floor of the exchange. For each transaction, the record contains the custom
er type, the trade
type, the broker's identification number, the number of contracts traded
, the buy-sell indicator
and the price. Most relevant, there are four different customer types for
each trade by each
floor trader: trading for their own account, trading for their clearing memb
ers' house account,
trading for another member present on the exchange floor, and trading
for any other type of
customer. For our purpose, only those trades related to dual trading, that
is, trades executed
5
for the trader's own account and for customers, are included in the sample.
We use a three-month window to examine the effect of the CME's rules. The sample
period covers May I through July 31, 1987 for the S&P 500 index futures contract with the
top-step rule effective from June 22. For the Yen, the sample period covers April I through
June 28, 1991, with the rule banning dual trading effective from May 20. The pre- and postrule samples are defined according to the event date from which the rule was imposed. For
both contracts, there are 35 days in the pre-rule sample and 29 days in the post-rule
sample. 10
To identify dual and other traders, we first calculate a trading ratio for each floor trader
for each day she is active. Specifically, define d
= (personal trading volume)/(personal
trading volume + customer trading volume), the proportion that personal trading volume is of
a floor trader's total trading volume on a day. For a floor trader, define a trading day as a
local day if d>0.98, a broker day if d<0.02 and a dual day if d lies on the closed interval
[0.02, 0.98]. 11
Based on these daily floor trader observations, we form several subsamples on which to
perform analysis.
Initially, we divide floor traders into a pre-rule and a post-rule sample.
Floor traders are categorized as dual traders, pure brokers or locals for each of the two
samples separately. A floor trader with at least one dual day in the sample is defined as a
dual trader. A floor trader with only local (broker) days in the sample is defined as a local
(pure broker). Since we intend to test for heterogeneity among dual traders, the dual trader
sample in the pre-rule period is split into four subsamples based on the observed occupational
choice of dual traders following restrictions. The four occupational choices are: to become a
6
pure broker, local, or dual trader; or to quit trading in the affected commodity. For example,
one subsample consists of those dual traders who became pure brokers following restrictions.
3. Trading Activity Before and After Dual Trading Restrictions.
3. a. Floor Trader Activity Before and After Dual Trading Restrictions
In this section, we describe the activity of floor traders in the two markets before and after
restrictions. These preliminary statistics offer some interesting contrasts between the floor
trader groups in the two futures pits. The number and activity level of dual traders were
reduced in both pits following restrictions. However, dual traders in the Yen pit spent less
time dual trading prior to restrictions and were affected more adversely following restrictions.
Following restrictions, activity by brokers and locals increased in the S&P 500 pit, and
decreased in the Yen pit.
The top half of table I contains summary statistics for the activities of floor traders in the
S&P 500 index futures. In the pre-rule period, there were 390 floor traders active on a given
day 12 , consisting of 210 locals, 26 pure brokers and 154 dual traders 13 • Following the topstep rule, the number of active floor traders increased while the number of active dual traders
dropped slightly", indicating that increased nondual trading more than offset the fall in
trading of dual traders. Also, in the pre-rule period, the average dual trader was more active
than the average floor trader, trading for more than 21 days out of the 35 sample days.
Almost 13 of these 21 days were spent dual trading, and 6 of 8 other days trading for their
o\1/n account. By contrast, following restrictions, the average dual trader spent just 5 of 21
active days dual trading, with the other days split almost equally between trading for
7
customers and trading for their own account.
The lower half of table I contains similar stati
stics for the Yen. There were I 08 active
floor traders in the pre-rule period, consisting
of 53 locals, 17 pure brokers and 38 dual
traders. Following restrictions, the number of
active floor traders fell. The number of activ
e
dual traders 15 also fell, and more sharply com
pared to the S&P 500 futures. Similar to the
S&P 500 futures, dual traders were more activ
e than nondual traders in the pre-rule period.
However, by comparison to the S&P 500 futu
res, dual traders in the Yen pit spent relatively
less time dual trading prior to the ban and trade
d mostly for customers following the ban.
3. b. Dual and Nondual Trading Days Befo
re and After Dual Trading Restrictions.
This section evaluates the typical daily trading
of a floor trader. We combine local days,
broker days, and dual days, independent of the
classification of the trader based on their
cumulative trading. These summary statistics
are shown in table 2. For both contracts, dual
trading days accounted for a significant portion
of pit activity, especially for customer trades.
The restrictions led to a sharp decline in the
amount of trading activity occurring on dual
trading days.
Statistics for the S&P 500 are presented in the
upper half of table 2. There were more
than 3.45 million contracts traded and 13,667
trader-days in the pre-rule period. Dual trading
days accounted for about 23% of all trader days
, 45% of all trades and 4 7% of total trading
volume. 72% of all customer volume was exec
uted by dual traders when they were dual
trading. Following the implementation of the
top-step rule, only 8% of trader days, 12% of
trades and less than 12% of trading volume occu
rred on dual trading days. Customer trades
8
on dual trading days fell sharply. As a result, the average number of daily trades and average
trading volume for customers fell from their pre-rule levels.
The lower half of table 2 shows statistics for the Japanese Yen. Almost 1.25 million
contracts were traded and there were 3,775 trader days in the pre-rule period. Relative to the
S&P 500, dual trading days were not a dominant part of total market activities prior to the
ban. Dual trading days accounted for 15% of trader days, 23% of trades and 25% of trading
volume. However, dual trading days still accounted for 41 % of all customer trading volume.
Following restrictions, dual trading days accounted for 6% or less of total trader days, number
of trades or trading volume.
4. Occupational Choice of Dual Traders Following Restrictions
4. a. Floor Trader Transition
In this section we follow dual traders in the respective markets from their behavior in the
pre-restriction period to their choice of occupation in the post-restriction period. A dual
trader has four possible reactions to the restrictions. First, they could continue to dual trade
according to the CME' s rules. As stated above, traders may still maintain an error account,
whose trades appear identical to proprietary trading in the data set. This error account
trading, combined with customer trading, will appear to be dual trading.
Also, they may dual
trade if they are not on the top step of the S&P 500, or if they switch from trading as a
broker to trading as a local once a day in the Japanese Yen pit. Second, they could become
locals. Third, they could become exclusive brokers. Fourth, they may simply exit the
particular contract which is subject to the restriction.
9
Panel A of table 3 reports this transition matrix for all floor traders around the
implementations of the respective restrictions. For our analysis, the most important numbers
are those which lie on the diagonal of the matrix. These indicate the number of floor traders
who continue in their original occupations following restrictions. For the S&P 500, 25% of
pure brokers, 67% of locals and 61 % of dual traders continue in their original occupation
after restrictions. The corresponding numbers for the Yen are 36%, 62% and 58%
respectively.
Many traders, especially pure brokers, traded only once during our pre-rule sample period.
To get a clearer picture of traders' choice, we report, in panel B of table 3, the transition
matrix for relatively active floor traders (defined as those who traded on at least 2 days during
the pre-rule sample period). For the S&P 500, 69% of pure brokers, 79% of locals and 64%
of dual traders continue in their original occupations after restrictions. The corresponding
numbers for the Yen are 68%, 86% and 63% respectively. Since we categorize a floor trader
as a dual trader if she traded just for one day in the 29 days following restrictions, whereas
brokers (locals) must trade for customers (themselves) every day, these numbers are strong
evidence that dual traders' occupational choice was primarily influenced by restrictions.
A surprising result is the large number of floor traders who quit trading in their home pit.
Panel A of table 3 shows that, for the S&P 500, more than 70% of pure brokers, almost 25%
of locals and 8% of dual traders were no longer active in the affected contract market for the
29 trading days following the top-step rule. For the Yen, the corresponding numbers are
63%, 36% and 12%, respectively. Since the normal attrition rate is I 0% or less per month
for floor traders in CME futures pits 16, the reported attrition rate for pure brokers and locals
10
here appears very high. Of floor traders switching occupations but not quitting, the primary
migration involves dual traders and brokers becoming locals.
Panel B of table 3, however, shows that the percent of discontinuing traders falls
dramatically for active floor traders, although (except for dual traders) the number is still
higher than the "normal" cutoff of 10%. For the S&P 500, 20% of pure brokers, 13% of
locals and 4% of dual traders were no longer active in their home pit following the top-step
rule. For the Yen, the corresponding numbers are 29%, 12% and 6%, respectively 17 •
4.b. Relative Trading Behavior of Dual Traders Prior to Restrictions
Table 4 shows summary trading statistics for separate categories of dual traders,
distinguished by their activity in the post-restriction period. Dual-locals (dual-brokers) are
those traders who were classified as dual traders in the pre-restriction period and chose to
execute exclusively personal (customer) trades in the post-restriction period. Dual-quitters are
those floor traders who were dual traders in the pre-restriction period and failed to trade in the
affected contract in the post-restriction period.
For both contracts, dual-brokers were predominantly involved in trading for their
customers on all their days in the pre-rule period. For example, in the S&P 500, dual-brokers
had only 24 local days out of 259 trader days. On their dual trading days, they traded on
average only 116.25 contracts for their own accounts, but 588.68 contracts for customers.
Similarly, for both contracts, dual-locals were almost entirely involved in trading for their
own accounts on all their days prior to the top-step rule. We observe a different trading
pattern for dual-quitters. In the S&P 500, they were primarily locals when they were not dual
11
trading, but mostly traded for customers on their dual days. In the Yen pit, discontinuing
dual traders traded mainly for customers on both their dual and nondual trading days.
The results in this section establish both the heterogeneity of dual traders in each market,
as well as the differential effect of restrictions on each type of dual trader. The evidence here
reveals that dual traders are heterogenous with respect to their trading patterns. In section 5,
we establish dual traders' heterogeneity with respect to their trading and execution skills.
5. Relative Skill Levels of Dual Traders Before Restrictions
5.a. Dual Traders' Personal Trading Skills
When dual trading becomes more costly, dual traders who are relatively skilled at trading
may trade only for their own accounts following restrictions while those who are relatively
less skilled may trade only for customers. Dual-quitters may be more or less skilled
compared to the average dual trader. We use the group of dual traders who dual trade for at
least one day in the 29 days following restrictions (referred to as "dual-duals") as the
benchmark group and compare their per contract median revenues with those of the other
three groups (i.e. those who switched to being locals or brokers, or quit). Since higher dual
trader revenues may be due to better information (obtained from their customers' trades)
rather than skill, the comparisons are made only for local days of dual traders prior to dual
trading restrictions.
Aggregate trading revenues for each dual trader are computed on a daily basis. For each
trader, and for each day, the value of purchases is subtracted from the value of sales, with
imbalances valued at the daily settlement price (marked-to-market). Daily revenues are then
12
divided by the nwnber of round-trip transactions for each floor trader, to obtain daily
revenues per contract.
Table 5 reports personal trading revenues of different groups of dual traders on their local
days prior to dual trading restrictions. The Wilcoxon Z statistic is used to test the null
hypothesis that the distribution of personal trading revenues of each dual trading group is no
different than that of dual-duals. Given the liquidity of these contracts, we expect per
contract revenues to be $25 (the minimum tick), or less. When the number of observations
(nwnber of trader days) in a group is relatively small, however, our calculated median
revenue values are much higher than $25. This observation is particularly valid for dualbrokers in the Yen pit, for which we have only two observations.
For both contracts, table 5 shows that dual-duals have close to the lowest median revenues
of the four groups. In the S&P 500, dual-duals had significantly lower per contract revenues
compared to the dual-quitters. Their median revenues are lower by a cash equivalent value of
$20.50 per contract, or about 80% of the minimwn tick of $25. For the Yen, there are no
significant differences between the median revenues of different groups of dual traders 18 •
These results suggest that, at least for the S&P 500, restrictions hurt relatively skilled dual
traders more than dual traders of average skill.
S. b. Dual Traders' Execution Skills
Similar to personal trading revenues, execution skills may be compared by calculating
average round-trip costs for customers. Asswning that floor traders are executing orders from
the same broad set of customers, any differences in trading costs for these customers across
13
floor traders can be attributed to the execution skills of the traders. 19
The customer costs are
calculated each day for each trader for all customer trades executed by
that trader. Customer
costs per contract are computed as the volume weighted average buy
price for each customer
minus the volume weighted average sale price (i.e., it is the opposite
of customer profits).
Several comparisons are performed both within each group and across
groups.
Table 6 reports the statistics for customer order execution costs. The
upper panel presents
the results for the S&P 500, and the lower panel presents results for the
Japanese Yen. For
each commodity, within each group, customer costs are compared on
the group's dual trading
and pure broker days. If dual traders are profiting from observing the
trading of their
informed customers, then on days when these traders dual trade, their
(informed) customer
costs will likely be lower than on days when the trader is only broker
ing. In other words, the
trader has an option on when to trade for her own account, and may be
exercising this option
when her informed customers are trading. A Wilcoxon z statistic is calcula
ted using the
trader day as the basic observation unit.
The test for the equality of customer costs20 on dual vs. broker days for
each group is
presented in the third row of each panel, labeled 'Dual vs. Broker Days'.
Only for the S&P
500 contract, for the dual-broker category, is the Wilcoxon z statistic
even marginally
significant. And, for this contract, customer costs are higher on dual trading
days--the opposite
of what should happen if dual traders' trades were informationally motiva
ted. These results
suggest that execution skill levels, rather than information, will be the
source of any
differences across groups.
In addition, and more related to our argument, each group's customer
costs are compared
14
to each of the other groups using only the broker days.
The Wilcoxon z statistic is
calculated to test the hypothesis that the respective dual trading group's customer execution
skill is no different than that of other dual trading groups. These statistics are presented in
the bottom three rows of each panel. For the S&P 500, two patterns stand out. First,
customers of dual-brokers had lower costs than the other groups. These costs are significantly
lower than those for customers of dual-duals (at the 5% level) and customers of dual-locals
(at the 10% level). The result indicates that customers did not lose the services of skilled
brokers following restrictions. Second, customers of dual-duals had the highest trading costs
among the four dual trader groups. Although the differences in costs are not significant, the
result is consistent 'with our earlier finding that restrictions affected relatively skilled dual
traders in the S&P 500 contract.
For the Japanese Yen contract, customers of dual-quitters had the highest execution costs
among all four dual trader groups. These costs are significantly higher (at the 5% level)
compared to customers of dual-locals. In fact, for both contracts, customers of dual-quitters
had the highest trading costs--with the exception of customers of dual-duals in the S&P 500.
These results suggest that dual-quitters were informationally motivated traders. Once the
information source (customers) disappeared (due to restrictions), their trading motive
disappeared too.
The results in this section provide evidence of differences in execution skills between
different dual trader groups. In particular, dual traders in the S&P 500 who became pure
brokers following restrictions had superior execution skills compared to two of the other three
dual trader groups. Dual traders in the Japanese Yen who quit trading following restrictions
15
had inferior execution skills compared to one of the other dual trader group.
Both these
results constitute good news for customers, and have implications for the effect
of the
restrictions on customer trading costs.
5.c. The Effect of Dual Trading Restrictions on Overall Customer Costs
In this section we examine the overall effect of dual trading restrictions on custom
er costs.
We combine all customer trading for all groups, and calculate customer costs
similar to the
procedure described in section 5.b.. These costs are calculated on a daily basis
for each day
in the sample, both before and after the dual trading restrictions.
To estimate the effect of
the restriction on customer costs, the following regression is estimated:
where, for day t, S, is the measure of customer trading costs (average buy price
minus
average sale price) in dollars, V, is customer trading volume,
VOL,
is the standard deviation of
buy prices for customer trades, M, is the number of floor traders trading for their
own
account, and D,
=
1 in the pre-rule periods and O otherwise. I-statistics are shown in
parentheses. N is the number of observations. This analysis parallels Smith
and Whaley
(1994) and Chang and Locke (1996).
If the restriction of dual trading increases customer costs, then a will be less than
zero.
4
Increased competition between floor traders should reduce customer costs, so
we expect a3 to
be negative. a 1 and a are expected to be positive, since customer costs tend to
increase with
2
16
volume and volatility. Results are presented in table 7. For neither contract is the coefficient
a4 significantly different from zero, indicating that customer costs were unaffected by
restrictions for both contracts. The other explanatory variables, with the exceptions of
volatility and volume (for the S&P 500), are insignificant. The coefficient on volatility is
positive and significant, as would be expected if marketrnakers widen the bid-ask spread when
price volatility increases.
6. Conclusion
Our aim was to show that trader heterogeneity can explain the conflicting empirical results
surrounding dual trading. Overall, the results confirm that there exists heterogeneity of trader
types on futures exchanges, so that regulatory restrictions such as a dual trading ban may have
disparate effects on different trader groups, and possibly unintended consequences for
customers. Those dual traders who discontinued dual trading in the S&P 500 index futures
had, on average, higher execution skills for both their personal and customer trading. For
example, dual traders who ceased trading the S&P 500 index futures had the highest personal
trading skills. However, on the whole, customers' transaction costs did not increase in the
S&P 500 following the introduction of the top-step rule. This may have been because dual
traders most skilled in executing for customers became exclusive brokers following the
restriction.
Our findings for the Japanese Yen are less pronounced. There is some evidence that dual
traders who quit trading the Japanese Yen had relatively poor customer execution skills.
Again, following the restriction customer transaction costs did not increase--perhaps because,
17
in this contract, the restriction did not appear to have hurt relatively skilled dual
traders more.
These results suggest that broad based measures, such as the bid-ask spread, or
contract
volume, are not necessarily sufficient to capture the effects of microstructure regulat
ion. The
seemingly divergent results in the dual trading literature may be due to the fact
that the
complexity of futures markets is more than allowed for by the typical microstructure
models.
18
REFERENCES
Chakravarty, S. (1994). "Should actively traded futures contracts come under the dual trading
ban?". Journal of Financial Intermediation, 14, 661-684.
Chakravarty, S. and Asani Sarkar (I 995). "The effect of broker competition on frontrunning
profits," Working Paper, Purdue University.
Chang, Eric C. and Peter R. Locke (1996). "The performance and market impact of dual
trading: CME Rule 552," Journal of Financial Intermediation, 5, 23-48.
Chang, Eric C., Locke, Peter R. and Steven C. Mann (I 994). "The effect of CME Rule 552
on dual traders," Journal of Futures Markets,14(4), 493-510.
Commodity Futures Trading Commission (1989). Economic analysis of dual trading on
Commodity Exchanges, Division of Economic Analysis, Washington, DC.
Fishman, Michael J. and Francis A. Longstaff (1992). "Dual trading in futures markets,"
Journal of Finance, 47(2), 643-71.
Grossman, Sanford J. (1989). "An economic analysis of dual trading", Rodney L. White
Center for Financial Research Paper 33-89, The Wharton School, University of
Pennsylvania.
Kuserk, Gregory, and Peter Locke (1992). "Scalper behavior on futures markets," Journal of
Financial Intermediation,
Locke, Peter, and P. C. Venkatesh ( I 996). "Futures market transactions costs," forthcoming,
Journal of Futures Markets, forthcoming, 1997.
Park, Hun and Asani Sarkar ( 1995). "Effect of dual trading on market depth in the S&P 500
futures market," Working Paper, University of Illinois at Champaign-Urbana.
Park, Hun, Sarkar, Asani and Lifan Wu (1995). "The costs of benefits of dual trading," Staff
Reports, Federal Reserve Bank of New York.
Roell, Ailsa (1990). "Dual capacity trading and the quality of the market", Journal of
Financial Intermediation, 1, 105-124.
Sarkar, Asani (1995). "Dual trading: Winners, losers and market impact," Journal of
Financial Intermediation.
19
Smith, Tom and Robert E. Whaley, I 994, "Assessing the
cost of regulation: The case of dual
trading", Journal of law and Economics, 37(1), 329-36.
Walsh, Michael J. and Stephen J. Dinehart, 1991, "Dual
trading and futures market liquidity:
An analysis of three Chicago Board of Trade contract
markets", Journal of Futures
Markets, I 1(5), 519-537.
·
20
Table 1
Activity by Floor Trader Types
S&P 500 and Japanese Yen Futures
Locals (brokers) refers to floor traders who traded exclusively for their own (customers) account during the sample period. Dual traders
refers to floor traders who traded both for their own and their customcn' accounu on
the
same
day at
least once during the sample
period. There arc 35 days before and 29 days after the dual trading restrictions for both contncts. The sample periods arc May t to July
31, 1987 for the S&P 500 futures and April 1 to June 28, 1991 for the Japanese Yen futures.
Local,
Number of:
Before
Dual Tr.id,rs
Brokers
After
Before
After
Before
All
After
Before
After
S&P 500
Traders
Trading days:
own account only
484
477
205
176
252
197
941
850
7,339
7,426
912
1,025
5,416
4,268
13,667
12,719
7,339
-7,426
1,595
1,563
8,934
8,989
648
1,668
1,560
2,693
3,173
1,037
3,173
1,037
21. ◄ 9
21.66
1 ◄ .52
14.96
6.33
7.93
9.49
10.58
2.57
8.47
1.66
3.17
12.59
5.26
3.37
1.22
154.74
147.17
390.49
438.59
45.57
53.9
255.26
309.97
18.51
57.52
44.57
92.86
90.66
35.76
90.66
35.76
912
customer only
1,025
dual
Active days per trader:
. own account only
15.16
15.57
15.16
15.57
4.45
4.45
customer only
5.82
5.82
dual
Active traders per day:
own account only
209.69
256.07
209.69
256.07
26.06
26.06
customer only
35.34
35.34
dual
Japanese Yen
Traders
Tr.ding days,
own account only
129
116
106
81
53
35
288
232
1,867
1,310
588
403
1,320
767
3,775
2,480
1,867
1,310
374
507
2,241
1,817
383
125
971
528
563
135
563
135
24.91
21.91
13.11
10.69
7.06
14.49
7.78
7.83
7.23
3.57
3.37
2.28
10.62
3.86
1.95
0.58
37.71
26.45
107.86
85.52
10.69
17.48
64.03
62.66
10.9◄
4.31
27.74
18.21
16.09
4.66
16.09
4.66
588
customer only
403
dual
Active days per trader:
own account only
14.47
11.29
14.47
11.29
5.55
5.55
customer only
4.98
4.98
dual
Active traders per day:
own account only
customer only
dual
53.34
45.17
53.34
45.17
16.8
16.8
13.9
13.9
Table 2
Activity By Trading Day Type
S&P 500 and Japanese Yen Futures
Local days (broker days) refers to trading days on whic~ floor traders tnded exclwively for their own (customers')
accounts. Dual tnding
days refers to trading days on which floor traders traded both for their own accounts and for their
customers. There are 35 days before
and 29 days after the dual trading restrictions for both contracts, The sample periods are May 1
to July 31, 1987 for the S&P 500 and
April 1 to June 28, 1991 for the Japanese Yen.
Local Days
Number of
Before
Broker Days
After
Before
Dual Trading Days
After
Before
All
After
Before
After
S&P 500
Trader days
8,934
8,989
1,560
2,693
3,17l
1,037
13,667
12,719
Transactions
483,428
513,677
8..,176
274,441
458,283
107,892
1,025,887
896,010
1,396,106
1,474,505
425,652
936,266
1,632,196
324,375
3,453,95◄
2,735,146
5 ◄ .11
57.15
53.96
101.91
14◄ .43
104.04
75.06
70. ◄ 5
54.11
57.15
42.76
39.78
45.l
43.63
Contract volume
Average daily trades:
own account
customer
Average daily volume:
156.27
164.03
own account
156.27
164.03
customer
Average trade size:
2.89
2.87
own account
2.89
2.87
customer
53.96
101.91
101.67
64.26
29.76
26.82
272.85
347.67
514.4
312.8
252.72
215.04
169.58
118.5◄
141.52
125.59
272.85
347.67
344.83
194.26
111.2
89.45
5.06
3.41
3.56
3.01
3.37
3.05
3.97
2.98
3.12
2.88
3.39
3.02
3.74
3.34
5.06
3.41
Japanese Yen
Trader days
2,241
1,435
971
910
563
1lS
3,ns
2,480
Transactions
128,910
64,407
65,298
49,215
59,399
7,261
253,607
120,883
Contnct. volume
558,528
297,661
374,240
339,488
309,749
36,166
1,242,517
673,315
57.52
44.88
67.25
54.08
105.5
53.79
67.18
48.74
57.52
44.88
25.86
11.59
38.01
26.6
Average daily trades:
own account
customer
Average daily volume
249.23
207.43,
own 21:count
249.23
207.43
customer
Average trade size
own account
customer
4.33
4.62
◄ .33
4.62
67.25
54.08
79.64
42.2
29.18
22.14
385.42
373.06
550.18
267.9
329.14
271.5
83.68
43.11
160.43
122.37
385.42
373.06
466.01
224.79
168.71
149.IJ
5.73
6.9
5.21
4.98
4.9
5.57
).24
3.72
4.22
4.6
5.85
5.JJ
5.78
6.74
5.73
6.9
Table 3
Floor Trader Transition
S&P 500 and Japanese Yen Futures
Floor traders are cla.ssified in a 35 day period before and a 29 day period aher dual trading restrictions by their
personal tnding volume as
a percentage of the swn of their cunomer .and personal trading volumes for trading days over the period.
8 of the active floor traders in
the S&P 500 futures and I active floor trader LO the Japanese Yen were dual traders before the restrictioru,
but tnded as pure brokers on
some days and as locals on other days following the restrictions. All of these traders are omitted from the sample.
Active Ooor traders,
used exclusively for the lower panel, are those who traded on at least two days in the pre-restriction period. Tbt
sample periods arc May
1 to July ll, 1987 for the S&P 500 and April I to June 28, 1991 for the Japanese Yen.
Post-restricti on choice for all floor traders
Pre-restriction
Pure broker
Local
Dual trader
choice
Discontinued
Pre-restrictio n total
S&P 500
Pure broker
52
8
5
l◄O
205
Local
7
325
36
116
48◄
Dual trader
14
61
1 ◄9
20
2 ◄4
Post-restriction
73
394
190
276
933
67
106
total
Japanese Yen
Pure broker
38
0
Local
Dual trader
Pon-restrictio n
total
80
2
46
129
4
12
30
6
52
◄3
92
33
119
287
Post-restriction choice for active Boor traders
Pre-restriction
choice
Pure broker
Local
Dual trader
Discontinued
Pre-restriction total
S&P 500
Pure broker
JI
Local
0
Dw1 trader
Pon-restrictio n
total
◄
9
45
305
32
51
388
I◄
61
1 ◄9
10
234
◄5
367
185
70
667
10
3◄
Japanese Yen
Pure broker
23
0
Local
0
71
2
10
83
Dual trader
4
II
30
)
◄8
Pon-restrictio n
total
27
82
33
23
165
Table 4
Activity of Dual Traders who Changed Occupations or Quit
S&P 500 and Japanese Yen Futures
Dual-brokers (dua1-locals) arc floor traders who were dassificd as dual traders
before the restrictions but switched to trading orJy for their
customers'(personal} accounu following the restrictions. Discontinued-duaJ
are door traders who were classuied :u dual traders before the
restrictions but quit trading in the affected contract or contract month
afterwards. The sample periods are May I to July 31, 1987 for the
S&P 500 and April 1 to June 28, 1991 for the Japancx Yen.
Dual-brokers
Before
T.-..ler Day Type
Local
Broker
Dual-locals
Aher
Dual
Broker
Before
Local
Dual-quinen
After
Before
Broker
Dual
Local
Local
Broker
Dual
S&P 500
T .-..ler days
24
102
135
204
892
JI
352
1,075
146
13
107
T ran.sactioru
127
6,318
18,798
18,108
57,347
567
36,062
59,433
5,805
234
9,582
Contract Volwne
345
32,344
95,166
72,260
165,705
1,786
151,524
166,580
17,685
1,701
23,963
Average daily trades:
5.29
61.94
139.24
88.76
64.29
18.29
102.45
55.29
39.76
18
89.55
own account
5.29
55.16
55.29
39.76
customer
A vcrage
daily volume:
own account
14.38
own account
2.72
120.29
88.76
317.1
704.93
354.22
116.25
588.68
354.22
5.12
5.06
3.99
6.13
5.12
4.89
185.77
18.29
47.29
57.61
430.47
154.96
121.13
233.69
154.96
121.13
185.77
317. I
2.72
customer
64.29
61.94
14.38
customer
Averagc trade size:
18.96
2.89
57.61
196.78
3.15
4.2
2.8
3.05
4.24
2.8
3.05
2.89
J.99
3.15
4.16
35.93
18
53.62
130.85
223.95
82.76
130.85
141.2
7.27
2.5
2.3
7.27
2.6)
Jap.ancsc Yen
Trader days
2
42
10
25
262
8
45
185
10
10
Tr.uuactions
19
31
6,550
760
2,196
17,070
31
3,163
8,278
110
515
1,112
Conuct volume
59
69,635
7,303
28,661
73,002
254
10,734
32,873
200
2,936
6,014
Average daily trades:
15.5
155.95
76
87.84
65.15
3.88
70.29
44.75
11
51.5
58.53
own account
15.5
62.76
44.75
11
customer
Average daily volume:
own account
29.5
own account
customer
I. 9
65.15
155.95
67.4
87.84
1,657.98
730.3
1,146.44
29.5
customer
Avenge trade size:
8.6
35.5
694.8
1,146.44
10.63
9.61
13.05
4.ll
10.63
10.31
7.53
31.75
238.53
177.69
20
224.49
177.69
20
278.63
1,657.98
1.9
278.63
3.88
4.28
31.75
14.04
8.19
3.39
3.97
1.82
3.58
3.97
1.82
4.28
13.05
8.19
1.86
12.95
51.5
45.58
293.6
316.53
41.11
293.6
275.42
5.7
5.41
3.17
5.7
6.04
Table 5
Dual Trader Personal Trading Revenues Per Contract on Their Exclusive
Local Days
S&P 500 and Japanese Yen Futures
Revenues per contract {in dollars) are calculated for dual traders' penoml trades on their local days for the pre-rule period. Dual-duals are
Door traders who dual tr2ded both before and after dual trading restrictions during our sample period. Dual-broken (dual-locals) are tloor
traders who dual traded before the restrictions but switched to trading only for their customers' (own) accounu following the restria.ions.
Oual-qu.inen are floor traders who dual traded before the restrictions but quit trading in the affected contract month afterwards. The z.
statistic tesu for differences in median revenues between continuing and the other groups of dual traders. Significant differences in median
values are starred. The sample periods are May 1 to June 20, 1987 for the S&P 500 and April 1 to May 19, 1991 for the Japanese Yen.
Dual-duals
Dual-brokers
Dual-locals
Dual-quiners
S&P 500
Mean Pro6u
0.43
259.5
19.5
158
Standard deviation
486.5
1121.5
477.5
680
Minimwn
-3565
-1075
-3240
-2624.5
-9.5
-217.5
-8.5
-23
17.5
fill
.!Z
~
50
84.5
42
175
3650
3308.5
3442.5
3712.5
32.5
-0.5
20.50*
0.47534
(0.6345)
0. 17173
(0.8636)
z-2.68627
(0.0072)
N-13
N-834
N-135
1st
Quartile
Medim
3,d Quartile
Maxirnwn
Difference in medians
Wilcoxon z·statistic
(p value)
N-464
Japanese Yen
Mean Profiu
41.25
275
-0.125
75
Sundud deviation
1617.5
318.75
1063.75
458.75
Minimum
-7412.5
50
-5487.5
-475
1st Quartile
-100
50
-137.5
-JOO
Medim
56.25
275
~
fill
3rd Quartile
193.75
500
226.25
625
Maximum
10500
500
4781.25
708.75
Difference in medians
218.75
20
-6.25
Wilcoxon z-statistic
0.8798
(0.379)
-0.74814
(0.4544)
z-0
(0.9999)
N-2
N-258
N-7
(p value)
N-89
Table 6
Dual Trade rs' Execution Skills
S&P 500 and Japanese Yen Futures
Trading costs (in dollars) are ca1culated daily for each trader's customers in the pre-rule period.
Dual-duals are floor traders who duaJ
traded both before and after duaJ trading restrictions during our sample period. Dual-broker
s (dual-locals) are floor traders who dual
uad~ before the resaictions but switched to trading only for their customers' (own) accounts following
the restrictions. Dual-quitters
are floor traders who dual lraded before the restrictions but quit trading ·in dlc affected contract montb
afterwards. Wilcoxon z statistics
for differences in median costs are in parentheses. p values arc given below. Signiftcant (at the 10%
level) z values are starred. The
sample periods are May I ro June 20, 1987 for the S&P 500 and April I ro May 19, 1991 ror
the Japanese Yen.
Dual-duals
Dual days
Broker
Dual-locals
Dual days
days
Dual-brokers
Broker
Dual days
Broker
days
Dual-quitters
Dual day
Broker
days
days
S&P 500 Index FubJ.res
Medlin costs
0
8.00
-11.25
-18.95
32.30
-42.85
-9.55
-6.70
Trader days
366
224
292
25
123
62
88
84
Wilcoxon z-statistic
(0.27)
0.7905
and p value
Dual
vs Broker days
(-0.21)
0.8334
0.1005
(-0.83)
0.3226
(-1.41)
0.1392
(-2.31)*
0.0176
(-0.89)
0.1261
(-1.23)*
0.095
(0. 76)
0.4895
(-1.64)
Wilcoxon z-statistic
and p value
On Broker Days:
vs. Dual-Dual
VS.
Dual-Local
vs. Dual-Broker
(0.18)
0.8367
1apanese Yen Futures
Median costs
18.75
31.25
59.375
9.375
21.875
3.125
37.5
139.06
Trader days
473
326
13
3
8
40
11
7
Wilcoxon z-statistic
and p value
Dual vs Broker days
(1.57)
0.1162
(-1.08)
0.2818
(-0.73)
0.4635
(1.09)
0.2771
(-0.58)
0.3904
(-0.42)
0.5536
(1.27)
0.1369
(-0.34)
0.6367
(2.33)*
0.0423
Wilcoxon z-statistic
and p value
On
Broker Days:
vs. Dual-Dual
vs. Dual-Local
vs. Dual-Broker
(1.75)
0.1294
Table 7
Customer Trading Costs Before and After Dual Trading Restrictions
S&P 500 and Japanese Yen Futures·
Changes in customer trading costs due to dual trading restrictions are estimated from the following regression:
S,
=
a0
+
a 1 Vr
+
0iVOL 1
+
a 3M,
-+-
a,Dr
+
e,
where, for day t , S is the measure of customer trading costs (average buy price minus average sale price) in dollars, V, is customer trading volume, VOL, is the standard deviation of buy prices for
1
customer trades, M, is the number of floor traders trading for their own account. and D, = 1 in the pre-rule periods and O otherwise. T-statistics are shown in parentheses. N is the number of
observations. The sample periods are May I to June 20, 1987 for the S&P 500 and April I to May 19, 1991, for the Japanese Yen.
S&P 500 Index Futures
Oo
a,
a,
a,
a,
20.21
(3.44)
--0.0005
(-2.00)
0.168
(4. 13)
--0.03
(--0.81)
4.74
(1.30)
N = 63
F = 11.083
Prob > F = 0.001
0.04
(1.25)
0.07
(0.079)
Japanese Yen
-1.62
(1.33)
-0.0001
(0.33)
1129.3
(2.47)
N = 63
F
Prob > F = 0.0015
=
5.036
Notes
1. The regulations allow affected exchanges to petition for relief based on 1) an acceptable audit
trail, or ability to track a floor traders' activities, or 2) a threat to the hedging utility and price
discovery function of futures markets, should the practice of dual trading be prohibited. All
affected exchanges have petitioned for relief, although the CFTC has yet to act on these petitions.
2. See below for a more detailed review of the theoretical literature.
3. In Fishman and Longstaff (1992), the effect of dual trading on liquidity may or may not be
negative. However, Sarkar ( 1995) shows that, if the assumption of fixed volume in the former
paper is relaxed, then dual trading may reduce liquidity. In Roell ( 1990), market liquidity is lower
because of dual trading, although some uninformed traders are better off.
4. We recognize these are idealized types. It is quite possible, for example, that trading skills are
necessary to exploit private information. We simply require that the primary source of dual
traders' revenues is trading skills in one case, and private information in the other. Our results
establish the existence of such types of dual traders.
5. CME's top-step rule (Rule 541) states: A member, who has executed an S&P 500 futures
contract order while on the top step of the S&P 500 futures pit, shall not thereafter on the same
day trade S&P 500 futures contracts for his account.
6. Rule 552 banned dual trading in all "mature liquid " contracts. The main criterion in
determining a "mature liquid" contracts was that contracts have "daily average volume of I 0,000
contracts or more ... over the previous six months" (CME Special Executive Report, May 3, 1991 ).
7. As of December 1991, five commodities had been affected by Rule 552: Pound Sterling, Swiss
Franc, Japanese Yen, Deutsche Mark, and Eurodollars. Our choice of the Yen was determined
by the availability of data. However, Chang and Locke's (1996) study of Rule 552 shows that
the Yen is representative of the group of affected contracts.
8. Locals are floor traders who trade exclusively for their own accounts.
9. Chakravarty (1994) shows that increased competition may lower the payoff to dual trading.
In his model, high volume markets will be more competitive, and less susceptible to dual trading
abuses. Chakravarty and Sarkar ( 1995) argue that the number of brokers in a pit and dual
traders' frontrunning profits are negatively correlated. Advances in trading technologies may also
make it easier to detect trading abuses.
IO. The difference in the size of the pre-rule and post-rule samples arises because the two
regulatory events do not fall exactly in the middle of our sample period.
11. The 2% filter is used to allow for the possibility of error trading. As Chang, Locke and Mann
(1994) state, "when a broker makes a mistake in executing a customer order, the trade is placed
into an error account as a trade for the broker's personal account. The broker may then offset the
error with a trade for the error account. A value of 2% for this error trading seems reasonable
from conversations with CFTC and exchange staff."
12. This number may be relatively high because all CME memberships allow for trading in the
S&P 500 futures pit.
13. We exclude floor traders who were locals on some days and exclusive brokers on other days.
There were 64 such traders in the S&P 500 futures pit and 16 in the Yen pit. This daily
switching is not considered dual trading, yet these traders are not necessarily brokers or locals.
14. Following the top-step rule, traders may still be identified as dual traders if they I) dual trade
off the top-step; or 2) are brokers who have a large percentage (more than 2%) of error trades
on their own account, making them appear as dual traders.
15. Following rule 552, traders may still be identified as dual traders if they I) switch from
trading as a broker to trading as a local once a day; or 2) are brokers with a large percentage of
error trades (more than 2%)) on their own account, making them appear like dual traders.
I 6. Traders enter and leave futures pits on a regular basis. The normal attrition rate refers to the
percentage of existing floor traders who discontinue trading in their home pit in the subsequent
month. The total number of traders in the pit need not fall, however, since new traders are
entering the pit each month. The I 0% number was obtained from informal conversations with
CME sources.
17. As suggested above, there are likely to be some exchange members who wander into these
pits infrequently, in addition to the regular, active floor traders semi-permanently stationed in the
pit. However, Kuserk and Locke (1992) present evidence of the lack of migration of traders
across various commodities within a day. Chang, Locke and Mann (1994) examine the exchangewide trading of traders in the currency and Eurodollar markets affected by Rule 552.
18. For the Yen, the two sample median test (with normal approximation) does show that dual
traders who became locals after dual trading restrictions had higher per contract revenues
compared to continuing dual traders. This result is significant at an 11 % level.
19. However, our data does not allow us to identify the end-users of the futures contract. In
other words, we know when a floor trader is executing an order for a customer, but we have no
information on the identity of that customer. Thus, floor trader-customer linkages, which are
likely important, are obscured.
20. Note that, since these are customer costs of trading, the expectation is that the numbers will
be positive for the typical broker. A negative number implies that customers are buying on
average at a lower price than they are selling, which is not consistent with the notion that
customers are demanding liquidity. However, the median customer costs are of both signs across
the groups and commodities.
