Liquidity, Transaction Costs, and Reintermediation in
Mary Jean and Frank P. Smeal Professor of Finance
Pennsylvania State University
∗ Ian Domowitz, 2001. Department of Finance, Smeal College of Business Administration,
Pennsylvania State University. The author may be contacted by email at email@example.com. This
paper was prepared for the Financial E-Commerce Conference of the Federal Reserve Bank of
New York, 2001. Several of the ideas presented here were motivated by stimulating discussions
with Ray Killian, of ITG, Inc. I have also benefitted from conversations with Mark Coppejans,
Jack Glen, Ananth Madhavan, and Benn Steil, and from their coauthorship with respect to pieces
of the puzzle, cited herein. Responsibility for how these pieces are put together, as well as for
errors and omissions, is mine alone, of course.
Liquidity, Transaction Costs, and Reintermediation in Electronic Markets
The central theme of this paper is the relationship between trading cost, technology,
and the nature of intermediation in the trading services industry. Electronic markets
embodying automated trade execution are linked to reductions in trading transaction
costs. Lower explicit costs are related to development and operating costs in an
electronic environment. Information available from electronic limit order books is
identified as a means of realizing implicit cost savings. The concept of liquidity
management in electronic environments is introduced, and its potential and use in
practice are illustrated using limit order book data. The empirical results suggest new
roles for brokerage and exchange operations, and competition between the two.
Reintermediation is defined as the reestablishment of a disintermediated institution, and
its nature is investigated for brokers and trading markets in an electronic environment.
Competitive advantage with respect to the provision of liquidity management services is
compared across types of reintermediaries.
The industrial organization of the trading services industry is changing rapidly. New
exchanges for the trading of financial instruments and novel forms of financial
intermediation appear on almost a continuous basis.1 Trading market structure is
marked by historically unprecedented market merger activity, as well as by consortia
and alliance formation in the provision of exchange services. Introduction of products is
joined by new methods of delivery and information transmission mechanisms.
Developments occur on a global basis, ignoring the demarcations of national borders.
Electronic financial markets lead the way in this evolution. The choice of
computerized trade execution systems, and many of the changes mentioned above, can
be explained through two simple observations:2
♦ Markets are firms, with the possibility of revenue sources distinct from those
obtained through market making services on the part of members.
♦ Markets are communications systems, and telecommunications is a better model
for exchange services than, say, the banking industry.
Network effects peculiar to communications systems are closely related to liquidity
and liquidity provision through changes in financial market structure. Although the link
between liquidity and trading costs is well known, the connection between automated
market structure and trading costs is not.3
The focus of this paper is on the potential of electronic markets to lower costs to the
trader, and implications of that potential with respect to the development of trading
markets. This perspective leads not only to a discussion of the future of exchanges, but
also of their new competition, novel forms of financial intermediaries.
Three questions motivate the analysis, contributing to a view of evolution in the
market for markets. Does trading in a venue characterized by automated trade
execution result in lower costs to the trader? What are the means, peculiar to electronic
1 The word, exchange, is loaded with regulatory implications; see, for example, Domowitz (1996).
For the purpose of this paper, exchange, market, and trading system are used interchangeably,
without reference to regulatory definitions.
2 This thesis is elaborated upon in Domowitz and Steil (1999), who link the choice of automated
market structure to cross-border trading, competition, market merger activity, and changes in
3 See Amihud and Mendelson (2000) for literature connecting liquidity and cost, and Domowitz
and Steil (1999) for studies of market quality across traditional and automated market structure.
market design, that permit the realization of cost savings? What do answers to these
questions imply about the nature of intermediation in an electronic environment?
Adoption of automated trading technology contributes to trading cost reductions. A
group of European automated exchanges, examined in section 2, exhibits costs in the
form of commissions and fees that are 41 percent below the global average. These
exchanges have implicit trading costs that are 66 percent below others, on average.
Trading costs across automated and traditional market structures also are investigated
by comparing executions via alternative trading systems and those accomplished
through Nasdaq OTC broker/dealers. Automated trade execution reduces total trading
costs by an estimated 32 percent over the sample period.
The way in which cost reductions might be obtained via automated trading systems
is the topic of section 3. Reduced explicit costs of trading can result from savings due to
lower development and operating costs of automated markets. Recent estimates of
exchange setup costs suggest that trading floor development is from two to forty times
as expensive as that for electronic marketplaces. Electronic order books reduce the costs
of market monitoring. They allow real-time assessment of liquidity, which suggests
active liquidity management to control implicit transactions costs. Data from a
computerized market suggest that observation of order depth through the book permits
larger trades in an environment characterized by more trading activity. Larger sizes may
be transacted with smaller spreads, conditional on order book information.
Liquidity management is illustrated by examining realized price impacts. Naïve order
submission strategies result in execution costs that are from 59 to 240 percent higher
than those observed in practice. Most notably, realized execution costs vary little as the
size of trade increases. Trading costs for trades done through the limit order book also
are less, on average, than costs incurred for "upstairs" negotiated trades done off the
system. The results are consistent with active liquidity management on the part of
market participants, and have interpretations with respect to order fragmentation
strategies and the economic benefits of off-exchange dealing.
Some implications of these results with respect to intermediation in the trading
services industry are addressed in section 4. Lower costs suggest disintermediation of
trading services, especially at the level of the brokerage function. The concept of
reintermediation is introduced, as the reestablishment of a disintermediated institution in
the new electronic environment. The issues revolve around product and services
provision, and the focus is on trade execution and investor decision support services.
Examples are provided, showing how brokerage reintermediaries may resemble
exchanges, and how electronic exchange reintermediaries may rebundle services to
coopt brokerage functions. Competitive advantages of each set of players are outlined,
in an attempt to gauge the future industrial organization of the industry. The paper
closes with a brief mention of the interaction between exchanges, brokers, the internet,
and the matching and search functions that are integral to trading activity.
Trading Costs and Electronic Markets
Trading Costs and
Interest in electronic markets is fueled by explosive growth in the use of automated
trade execution in practice. Outside the U.S. and a few emerging markets, there are
few equity or exchange-traded derivatives trading systems that are not fully, or at least
partially, automated.4 Foreign exchange trading is increasingly done through automated
systems, such as EBS and Reuters 2002. The number of initiatives in the fixed income
arena has been estimated to be as large as 60 on a global basis, represented by
eSpeed, Euro MTS, BondLink, BondConnect, and BondNet.
It also is not difficult to motivate the potential importance of trading transaction
costs. Consider, for example, an equally weighted global portfolio of stocks.5 Over
1996:3 through 1998:3, one-way total trading costs for this portfolio average 71 basis
points (bps). If the portfolio turns over twice a year, 285 bps in total costs are incurred.
Average annual portfolio return over the period is 1228 bps. On this basis, trading costs
alone account for 23 percent of returns.
The issue addressed in this section is the historical relationship between these items
of interest. The empirical link between automated markets and trading costs is first
examined on an international level, then with respect to finer information on trading in
2.1 International evidence
2.1 International evidenc
4 Some partial automation simply represents gradual changeover to fully electronic operations.
The London Financial Futures and Options Exchange, for example, began such a transition in
November, 1998, closing the floor completely on November 27, 2000.
5 See Domowitz, Glen, and Madhavan (2000) for discussion, analysis, and precise definitions of
cost. Data are based on institutional trades representing the activity of 135 institutional investors
in 42 countries. These institutions accounted for 28 billion shares in 632,547 trades, using 700
global managers and 1,000 brokers. Transactions costs used here include both explicit costs in
the form of fees and commissions, as well as implicit costs, such as the bid-ask spread.
Trading transactions costs are falling world-wide, illustrated by the regional declines
in Figure 1. Over the 1996:3 through 1998:3 period, costs declined by 10 to 53 percent
by region, averaging about 16 percent. Explanations include competition for order flow,
shifts of trading strategies to accommodate liquidity differences, more institutional
trading, and pressure from new trading systems and regulatory authorities.6
Adoption of automated trading technology contributes to trading cost reductions.
This is illustrated in Figure 2, which contains explicit and implicit cost figures for several
computerized equity markets, in comparison to an overall average of transactions costs
over 42 countries.7 Implicit costs for the countries shown range from 5 to 15 bps, one
way. The global average is 25 bps, and the U.S. number is approximately 30 bps.
Sweden, for example, exhibits implicit costs which are 62 percent below the global
figure, accompanied by explicit costs which are 43 percent below other countries, on
The setup and operating cost advantages of an electronic trading venue suggest
that commissions and fees might be lower in automated markets. Domowitz and Steil
(1999) document such savings from the perspective of the exchange. Automated and
floor system development plans indicate that the latter are far more costly, for example.
Recent estimates of automated system development are from $10 million to roughly
$100 million. Floor development costs can range from approximately $200 million to
$400 million.8 Annual operating cost savings are a bit more anecdotal, but some
estimates put them at 40 to 60 percent of human resource and ancillary service cost.9
Although lower explicit costs are observed in automated venues, the proposition that
automated market structure systematically reduces such costs is not universal.
Australia's explicit costs are 10 percent above the global average. More strikingly, they
6 Competition between markets for international order flow is increasing (Foerster and Karolyi
(1999)), which can reduce domestic market spreads (Domowitz, Glen, and Madhavan (1998)).
Domowitz and Steil (1999) discuss the competitive landscape of exchange services competition.
An example of regulatory pressure, and its effects, is given by Barclay, Christie, Harris, Kandel,
and Schultz (1999).
7 See also Pagano and Röell (1990), Pirrong (1996), and Schack (1999) for evidence on reduction
of spreads in automated limit order books, relative to dealer and floor trading venues.
8 The London Stock Exchange and Deutsche Börse expended over $100 million implementing
Sets and Xetra, while Tradepoint was developed for under $10 million. LIFFE’s floor development
plan, abandoned in the spring of 1998, was priced at over $400 million. A bond futures trading
floor project in Chicago was completed in 1997 at a cost of approximately $200 million.
9 Sydney Futures Exchange and Toronto Stock Exchange, respectively.
are 83 percent above the average of the other three countries employing electronic
markets in figure 2.
Higher explicit costs may be incurred for industry and regulatory reasons having little
to do with the automation of trading market microstructure, however. Such effects are
particularly prominent in emerging equity markets.10 In derivatives markets, higher
explicit costs are sometimes associated with the nature of the contracts traded. On the
London International Financial Futures Exchange's CONNECT system, transaction fees
on commodities such as sugar, coffee, and cocoa are 62 pence per lot per side,
compared to 25 pence for financial futures contracts. The argument given by the
exchange is that more costs are incurred for commodities, given infrastucture that
includes expenses for grading and delivery systems.11
2.2 Evidence from the U.S.
In the the U.S., trading costs may be compared by examining executions via
electronic communications networks (ECNs) and those accomplished through traditional
brokerage and markets. Domowitz and Steil (1999,2001) carry out such studies using
single-institution trade data between 1992 and 1996. The institutional investor was the
largest single user of ECNs among active money managers during the period. The
investor had no soft commission arrangements with brokers as a matter of company
policy, thus making its trade data well suited for cross-market comparisons.12 The data
used in what follows are the same as employed by Domowitz and Steil (2001) with
respect to the OTC market. Total trading costs include price impact, determined as a
geometric average of realized and effective spreads, and measured relative to short-run
Figure 3 contains figures on trading cost savings on OTC trades by year, from 1992
through 1996. The numbers are calculated by volume-averaging total trading costs
incurred through executions by 4 automated trade execution facilities and 34 traditional
10 The correlation between explicit costs and whether equity trades take place in an emerging
market is 0.47, given in Domowitz, Glen, and Madhavan (2000).
11 See "Fee Outcry as Connect Goes Live," Financial Times (November 24, 2000), p. 17.
12 Soft commission trades are those allocated not because a broker offers the lowest cost of
execution on a given transaction, but because the institutional investor is obliged to pay a
minimum level of annual commissions to the broker in return for services unrelated to trade
13 Details are available in Domowitz and Steil (1999); the cost measures were constructed by SEI,
broker/dealer operations. The savings reported then are computed as the difference
between automated execution costs and those of the traditional dealers, relative to the
traditional dealer average cost.
Average savings from using automated trade execution over the full sample are 32.5
percent, relative to traditional broker/dealer operations. Savings using automated
systems are evident for every year except 1995, and range from 0.001 percent to 67
percent. The 1995 exception stems from a single outlier in the data. During the first
half of 1995, a 37 percent cost was recorded on a trade done through Instinet, a system
that averaged a total cost, as percent of value traded, of 0.55 percent over the 5-year
Trading costs depend on the difficulty of the trade, of course. Trade difficulty is
obvious when viewed from the trading desk, but hard to quantify. Figure 4 contains
cost savings disaggregated by various market indicators, all of which have been used
previously in the measurement of economic characteristics of trading costs.14 These
characteristics include shares per trade, market capitalization of the stock, market beta
of the stock, annualized daily standard deviation of returns for the traded issue, and the
inverse of the share price. Execution costs may diminish with firm size, owing to
relatively better liquidity and reduced informational asymmetries. Larger trades are
more difficult, hence more costly, possibly due to larger inventory costs in intermediated
settings or because of information content. Costs rise with volatility, especially in
intermediated venues, given some degree of risk aversion. Trading costs are related to
price levels, another measure of company size.
Savings from automated execution are evident even for "difficult" trades, measured
as having above median values of trade size, beta, and volatility, as well as below
average market capitalization. Such savings are observed regardless of price level.
Although conventional wisdom that "easier" trades are more often sent to automated
systems is born out by the data, it is still the case that automated systems afford
savings even in more difficult trading situations.
Some examples from figure 4 illustrate the magnitudes involved. Automated systems
are almost 60 percent less expensive than their traditional counterparts in relatively high
volatility environments. A similar estimate is obtained for high volatility trades as
14 See, for example, Bessembinder and Kaufman (1997) and Harris (1994).
measured by market beta. For small trades, the cost savings are over 28 percent, and
for large trades, automated systems still deliver a 10 percent cost savings. For large
capitalization stocks, there is little difference between automated systems and traditional
dealer operations. Cost savings on the order of 85 percent, relative to dealers, are
available through automated execution for small stocks, however. Such indicators
cannot capture all aspects of trade difficulty, and trading costs certainly embody other
factors.15 Nevertheless, the results remain strongly suggestive of savings due to
automation of trade execution technology.
r Books and Strate
r Books and Strate
Smaller explicit costs can result from development and operating cost advantages of
automated markets, passed on to the market's customers. It is not so obvious that
automated market structure should translate directly into lower implicit trading costs. A
claim to that effect is reminiscent of assertions that some market structures offer more
liquidity than others, all other things equal. Market structure does not create liquidity;
traders do. Similarly, automation by itself cannot induce lower implicit costs of trading.
On the other hand, a contribution of the theoretical and experimental literature on
auctions is the observation that the form of the trading institution matters. Auction
design affects trader behavior, the properties of transactions prices, and market
efficiency. This set of findings also is found in the related literature on financial market
The practical issue then is to identify the means by which electronic market design
allows traders to reduce price impact costs. Abstracting from market structure
considerations, how might traders accomplish such a reduction?
Suppose liquidity were to vary over the course of a trading day. Monitoring of
liquidity characteristics permits what might be called strategic liquidity management, the
goal of which is to reduce transactions costs.16 In the theoretical literature, monitoring
possibilities introduce discretionary timing of trades.17 The competitive behavior of
discretionary traders leads to trading in the lowest cost period presented by the market.
15 So-called "opportunity cost" is an example, a concept which has to date eluded effective
empirical measurement, regardless of study.
16 One might also add the management of risk associated with liquidity risk in the context of
more general portfolio problems.
17 See Spiegel and Subrahmanyam (1995), Admati and Pfleiderer (1988). and Scharfstein and
This, in turn, induces clustering in liquidity supply and demand, reinforcing the time-
varying nature of liquidity provision.
One mechanism through which such management may be accomplished is the
electronic limit order book, a feature of most automated market structures.18 The
existence and dissemination of order book information sharply reduces the costs of
monitoring the market, and permits real-time assessment of liquidity.
The potential for liquidity management is investigated in Coppejans, Domowitz, and
Madhavan (2000). The authors examine liquidity dynamics and cost control strategies
using data from the OM market for Swedish stock index futures contracts.19 Liquidity in
that paper is characterized by depth of market at different numbers of ticks away from
the quote midpoint. Another possible measure is the size-adjusted spread, defined as
the round-trip spread cost of doing a trade as a function of trade size. These liquidity
characteristics are observable to traders through the electronic book.20
Figures 5 and 6 illustrate the possibilities. As depth increases or the size-adjusted
spread decreases, the number of trades in a five-minute period goes up and the average
size of a trade increases. Trading frequency rises by 53 percent as depth goes from
below median to the 95th percentile, for example. The average size of trade rises by 13
Larger trades in an environment characterized by more trading activity are possible,
as a function of observable depth in the order book. Further, larger sizes may be
transacted with smaller spreads, by monitoring the book. Traders consume liquidity
when it is plentiful. They supply liquidity when needed.21
Traders appear to be using the electronic system to monitor liquidity and
strategically trade against the book, with respect to cost control. This is evident from
figure 7, which contains price impact calculations. The columns that increase in size
with trade size are price impacts for each size, under a naïve trading strategy that
18 See, for example, Domowitz (1993).
19 The database comprises the complete limit order book for Swedish stock index futures
contracts from the period 7/31/95 through 2/23/96. Information is time-stamped to the second.
Transactions files and order information are matched. The order book is reconstructed from the
raw data and completely consistent with transactions reported. "Off-exchange" cross trades are
isolated and matched to the order book files.
20 See also Irvine, Benston, and Kandel (2000) for alternatives.
- Internet Banking: Developments and Prospects
- Electronic Finance: Reshaping the Financial Landscape Around the World
- Electronic Communication Networks and Liquidity on the Nasdaq
- Impacts of the Internet on Financial Services