EWE21 7/24/03 9:05 PM Page 469
`
`Chapter 21
`
`New Electronic Trading
`Systems in Foreign
`Exchange Markets
`
`Dagfinn Rime
`Norges Bank and Stockholm Institute for Financial Research
`
`I. Introduction
`II. The Structure of Foreign
`Exchange Markets
`A. Information and Agents
`B. Institutions
`C. Interbank Trading Options
`D. Transparency
`III. Direct Trading and Voice Brokers
`A. Dealer Behavior
`IV. Electronic Brokers
`A. Transparency
`B. Liquidity
`C. Transaction Costs
`
`D. The Future of Direct Trading
`and Voice Brokers
`E. Policy Implications
`V. Internet Trading
`A. The Emergence of Nonbank
`Customer Trading
`B. Internet Trading with Banks
`C. Possible Scenarios
`VI. Summary
`A. Web Sites on Trading and
`Networks
`References and Further Reading
`
`The foreign exchange market can be divided in two segments: the
`interbank market and the customer market. Two advances in trading
`technology, electronic brokers in the interbank market and internet
`trading for customers, have significantly changed the structure of the
`foreign exchange market. In this chapter, we explain the functioning
`of electronic brokers and internet trading and discuss the economic
`consequences. © 2003, Elsevier Science (USA).
`
`New Economy Handbook
`Copyright 2003, Elsevier Science (USA). All rights reserved.
`
`469
`
`GAIN CAPITAL - EXHIBIT 1014
`
`

`

`EWE21 7/24/03 9:05 PM Page 470
`
`470
`
`Rime
`
`Bid–ask spread Difference between the best buy price (ask) and best sell
`price (bid). The initiator of a trade buys at the ask and sells at the lower
`bid price. The spread is a measure of transaction costs. The buy price is
`also called the “offer.”
`Broker Brokers match dealers in the interbank market without being
`a party to the transactions themselves and without taking positions
`(cf. dealer).
`Call market A market where all traders trade at the same time when
`called upon.
`Counterparty credit risk The risk that the market participant on the other
`side of a transaction will default. Due to the large trade sizes in foreign
`exchange markets, credit risk is an important issue.
`Dealer A person employed by a bank whose primary business is entering
`into transactions on both sides of wholesale financial markets and
`seeking profits by taking risks in these markets (cf. broker).
`Dealer market Market where orders for execution pass to an inter-
`mediary (dealer) for execution.
`Interbank market The market where dealers trade exclusively with each
`other, either bilaterally or through brokers.
`Limit order Order to buy a specified quantity up to a maximum price or
`sell subject to a minimum price (cf. market order).
`Liquidity Characteristic of a market where transactions do not exces-
`sively move prices. It is also easy to have a trade effected quickly
`without a long search for counterparties (“immediacy”). Liquid markets
`usually have low bid–ask spreads, high volume, and (relatively) low
`volatility.
`Market maker Dealer ready to quote buy and sell prices upon request.
`The market maker provides immediacy (liquidity services) to the market
`and receives compensation through the spread. There is no formal oblig-
`ation to quote tight spreads; rather, market making is governed by
`reciprocity.
`Market order Order to buy (or sell) a specified quantity at the best pre-
`vailing price (cf. limit order).
`Order-driven market Market where prices are determined by an order
`execution algorithm from participants sending firm buy and sell orders,
`which are incorporated into the limit order book (cf. quote-driven or
`dealer market).
`Order flow Signed flow of transactions. The transaction is given a
`positive (negative) sign if the initiator of the transactions is buying
`(selling).
`Price discovery Determination of prices in a market. Incorporation of
`information into prices.
`
`

`

`EWE21 7/24/03 9:06 PM Page 471
`
`21. E-Trading Systems in Foreign Exchange Markets
`
`471
`
`Quote-driven market Refers to a market where market makers post bid
`and ask quotes upon bilateral request. In the interbank market, these
`prices are on a take-it-or-leave-it basis (cf. order-driven market).
`Transparency Ability of market participants to observe trade information
`in a timely fashion.
`
`I. INTRODUCTION
`
`The 1990s gave us what might prove to be the two biggest changes in
`foreign exchange market structure since World War II: electronic brokers
`were introduced into the interbank market in 1992, and in the late 1990s
`the Internet became available as a trading channel for customers. What are
`the consequences for the market of these innovations? Is there any reason
`to believe that these technological developments have influenced the
`market in any significant way? Do not dealers in the foreign exchange
`market still fulfill their function as liquidity providers and aggregate infor-
`mation in their price setting? And, do not basic macroeconomic variables
`still drive exchange rates, irrespective of trading technology?
`In an ideal world with perfect information, these changes to the institu-
`tions of trading probably would not matter that much at the macro-
`economic level. In such a world, exchange rates would be determined by
`expectations regarding macroeconomic fundamentals like inflation, pro-
`ductivity growth, and interest rates. Exchange rates will be efficient asset
`prices when all market participants observe these fundamentals and agree
`on how they influence exchange rates. Furthermore, provision of liquidity
`would be much less risky than in a situation with imperfect information.
`However, as empirical evidence has shown all too clearly, models of an ideal
`world with perfect information do not hold, at least not for horizons shorter
`than a year or so.
`The microstructure approach to foreign exchange has made some
`promising steps toward solving some of these puzzles (see Lyons, 2001a).
`This approach differs from the traditional macroeconomic approach by
`allowing for imperfect information and heterogeneous agents and, thereby,
`leaving a role for trading institutions as such. In such a world, technologi-
`cal changes such as the introduction of electronic brokers and Internet
`trading may be significant because they change the structure of the market.
`A different market structure changes the game played between the
`market participants. This may influence information aggregation capabili-
`ties and incentives for liquidity provision and, thereby, different aspects of
`market quality like efficiency (price discovery), liquidity, and transac-
`
`

`

`EWE21 7/24/03 9:06 PM Page 472
`
`472
`
`Rime
`
`tion costs. We are interested in understanding market structure because
`a well-functioning foreign exchange market is important for the macro-
`economy. This chapter considers the impact of technological advances on
`the foreign exchange market by focusing on these properties of market
`quality.
`The new economy and foreign exchange markets is a vast subject. We
`limit ourselves to the two major innovations in trading technology because
`trading institutions are an important part of a financial market’s structure.
`Furthermore, several studies show that trading is important for the
`determination of exchange rates. There is particular focus on a property
`of market structure called transparency, i.e., how much of the trading
`process market participants can observe. Because trading is an important
`determinant of exchange rates, observation of the trading process is
`important to enable dealers to set the “correct” exchange rates. On a more
`general level, transparency relates to how efficiently dealers can aggregate
`information.
`There are of course many other uses of information and communication
`technology (ICT) that have obviously influenced the markets that we do
`not address here. These include information providers such as Reuters
`and Bloomberg, computers’ calculation capabilities and the importance
`for option trading, and of course network technologies and computers in
`general. Two other technological innovations deserving special atten-
`tion that we do not consider are the newly started settlement service
`CLS Bank (Continuous Linked Settlement), which went live on September
`9, 2002, and the netting technology FXNet. The former links all participat-
`ing countries’ payment systems for real-time settlement. With such a
`system in place in 1974, the famous Bankhaus Herstatt default would
`never had happened. FXNet is a technology for netting out gross lia-
`bilities. Both are very important for the handling of counterparty credit
`risk.
`Sections II and III provide the background for the introduction of elec-
`tronic brokers and Internet trading. A brief description and history are
`given of the structure of the market prior to these innovations, followed by
`some considerations that dealers take into account in their trading. The
`trading institutions of the 1980s are referenced to clarify the differences.
`Section IV discusses electronic brokers, whereas Section V discusses Inter-
`net trading. Section VI provides a summary.
`
`

`

`EWE21 7/24/03 9:06 PM Page 473
`
`21. E-Trading Systems in Foreign Exchange Markets
`
`473
`
`II. THE STRUCTURE OF FOREIGN EXCHANGE
`MARKETS
`
`Before we discuss electronic brokers and Internet trading, we need an
`overview of the general structure of the foreign exchange market so as to
`be able to understand the impact of these new trading institutions.Although
`electronic brokers were undoubtedly the most significant structural change
`in the 1990s, the general description given here is valid for the structure
`both before and after the introduction of electronic brokers. The reason is
`that brokers were present in the market before electronic brokers were
`introduced. The introduction of Internet trading, on the other hand, is still
`very recent, but it may prove to be the most significant structural shift of
`the first decade of the twenty-first century. This shift has the potential to
`overthrow the general structure of the market completely, a point that we
`come back to in Section V.
`
`A. INFORMATION AND AGENTS
`
`The foreign exchange market is the oldest and largest financial market
`in the world, with $1200 billion changing hands every day (April 2001).1
`These trades can be divided into interbank trades and customer trades, rep-
`resenting the two segments of the market. In the interbank market, trading
`is either direct (bilateral or taking place between dealers) or brokered
`(interdealer trades). Prior to the advent of the Internet, customers traded
`only with banks. We could have added customer-to-customer Internet-
`based trading sites, but we feel it is too early to pay them the same atten-
`tion as the three methods already mentioned (interbank, both direct and
`brokered, and customer–bank). In the 1990s, the market was often divided
`into three groups: customers, dealers, and brokers. However, as brokering
`becomes more and more electronic and also is open to customers through
`the Internet, we feel that it is more natural to focus on two main groups of
`traders: customers and dealers. The customers are the ultimate end-users of
`currency, and they typically make the largest single trades. Customers may
`be central banks, governments, importers and exporters of goods, and finan-
`cial institutions like hedge funds.
`Important characteristics of the foreign exchange market are that cus-
`tomers do not have access to the interbank market and that they do not
`trade with each other (except on the customer-to-customer sites mentioned
`
`1This number includes spot, forward, and swap volumes. In the following, we will focus on
`spot trading because spot is the most fundamental.
`
`

`

`EWE21 7/24/03 9:06 PM Page 474
`
`474
`
`Rime
`
`earlier, which we return to later). The trading that takes place with cus-
`tomers is private information for the banks, and dealers stress the impor-
`tance of seeing customer flows. An interesting question is what kind of
`information this trading may reveal. For understanding the concept of infor-
`mation in the foreign exchange market, we need to add some details to what
`we mean by information, and Lyons (2002) suggests the description given
`in Table 1. The starting point is the expression of an asset price as the dis-
`counted expected value. Information may concern the expected value, the
`payoff-relevant part, or the discount rate (including the risk premium).
`In the upper left corner of Table 1, concentrated payoff-relevant infor-
`mation, or information on risk-neutral valuations, is the kind of private
`information that is typical in equity markets. In the case of foreign
`exchange,
`it probably does not constitute the main motivation for
`information-based trading. Changes in central banks’ interest rates are too
`infrequent and too shrouded in secrecy for private information about these
`rates to be a major driver of trading. However, private information about
`interventions is a possible candidate because central banks sometimes
`perform their interventions through particular banks. Bettina Peiers found
`that the exchange rate changes made by Deutsche Bank were leading the
`rates of other banks through rumors of interventions.
`Rather, Lyons (2002) suggests that it is the lower row in Table 1 that is
`most relevant in foreign exchange markets. The information that needs to
`be aggregated by the market is not concentrated on a few people, but rather
`dispersed among many. In their 2002 paper, Martin D. D. Evans and Lyons
`present a model where customer trading represents portfolio shifts and
`signals changes in risk premiums (information of the type in the lower right
`cell of Table 1). A risk premium arises as the compensation necessary to
`induce others to take the other side of the portfolio shift. This risk premium
`must be a permanent change in exchange rates because the new holders of
`currency must want to continue holding the currency.Their empirical results
`confirm this.
`
`Table 1
`Sources of Private Informationa
`
`Payoffs
`
`Discount rates
`
`Concentrated
`Dispersed
`
`Interest rate changes, interventions
`Expectations, information interpretation
`(“market sentiment”)
`
`Risk premiums (risk aversion,
`portfolio shift)
`
`aLyons (2002) suggests that information in the upper left corner is unlikely in foreign exchange
`markets and that the lower right corner is the more relevant.
`
`

`

`EWE21 7/24/03 9:06 PM Page 475
`
`21. E-Trading Systems in Foreign Exchange Markets
`
`475
`
`We can also make an argument for the lower left cell. Because no model
`so far has proven to be the correct model for exchange rates, agents are
`likely to have different views on exchange rates. Customers’ trading counts
`the “votes” of the customers on where the exchange rate is heading. The
`hypothesis that information is dispersed has implications for optimal
`market structure: what kind of trading institutions are best fit for aggre-
`gating dispersed information? We come back to this in Section IV.
`
`B. INSTITUTIONS
`
`The market structure has not always been like the present structure.
`Prior to telecommunications, trading in foreign exchange markets could be
`described as a centralized call market. Trading in foreign exchange can be
`traced back to ancient times, when foreign exchange trading was a way to
`circumvent the ban on usury. In the sixteenth century, trading in foreign
`exchange occurred every third month at fairs in the Genoa area, each of
`which lasted for 8 days.2 However, telecommunications changed the general
`structure of the foreign exchange market, and it has been more or less unal-
`tered from the early 1930s up to the present.3 Drawing on the theory of the
`microstructure of financial markets, we can describe the current interbank
`foreign exchange market as follows:
`
`1. Trading is decentralized across several locations, as opposed to
`centralized on an exchange as is the case in many equity markets.
`2. There is continuous trading around the clock, as opposed to only
`when called upon as in a call market.
`3. There are several dealers that provide liquidity, as opposed to the
`specialist on the NYSE floor in earlier days, for example.
`4. Liquidity is both quote-driven, i.e., created by quoting bid and ask
`prices in response to trading initiatives (market making or dealer
`market), and order-driven, i.e., by entering limit orders with brokers
`(auction market).
`5. The market is relatively opaque, i.e., has low transparency compared
`with many equity markets.
`
`2The Genoa fairs lasted from 1532 to 1763 [see chapters by M. de Cecco in Newman et al.
`(1992)]. After this, Amsterdam and then London took over as the main location for currency
`trading.
`3It is difficult to find information on market structure from earlier periods. Several brokers
`started up in London in the 1930s with brokering over the telephone. Really active trading in
`foreign exchange, however, did not start until the mid-1960s, when regulations were eased in
`several European countries.
`
`

`

`EWE21 7/24/03 9:06 PM Page 476
`
`476
`
`Rime
`
`The introduction of telecommunications allowed decentralized trade of
`the asset foreign exchange, as is most natural. Banks want to be present
`where the customers are, and because an exchange rate is the relative price
`of two assets from two different countries, it is natural to have a decen-
`tralized market. Given that customers are in different time zones and may
`have an interest in the same asset, say $, trading must also be continuous
`around the clock. Finally, given the geographical pattern of customers and
`the fact that several banks serve them, it is natural to have a number of
`dealers acting as liquidity providers in each currency pair. The decentral-
`ized structure also makes it very difficult to regulate foreign exchange
`trading, and the market structure has therefore evolved endogenously.
`These factors, together with the lack of regulatory disclosure requirements,
`mean that the foreign exchange markets are characterized by low
`transparency.
`All of this has economic consequences. Low transparency means that few
`of the dispersed signals that order flow may reveal will be observed by a
`single dealer. In a centralized call market, which is more transparent, infor-
`mation aggregation will typically be faster and more accurate. The lack of
`regulation is also important. Disclosure requirements are imposed on
`equity markets so that more trading, and hence more information, is
`observed by the market participants. As will be discussed later, the trading
`institutions also have implications for risk sharing.
`
`C. INTERBANK TRADING OPTIONS
`
`Foreign exchange trading typically follows a sequence. Customers’
`trading is the primary source of currency demand, and the sequence starts
`with a customer contacting her bank with a wish to trade (dealers never
`take the initiative). The bank acts as market maker and gives quotes to the
`customer. Customers do not have access to the interbank market, so an
`exporter cannot contact his counterparty or the counterparty’s bank
`directly. For a customer, trading with the counterparty directly involves
`credit risk, which could be handled more efficiently by a bank. A dealer in
`the bank then turns to the interbank market to cover the customer trade.
`Interbank trading actually accounts for between 60% and 80% of the total
`volume of foreign exchange trading, and we will come back to this issue in
`the next section because it is closely related to trading institutions.
`In the interbank market, the dealer has several options, as illustrated by
`the 2 ¥ 2 matrix in Table 2. In a multiple-dealer market, the dealer may
`choose to provide liquidity as a market maker and give quotes when con-
`tacted by other dealers (incoming trade), or he may trade on other market
`
`

`

`EWE21 7/24/03 9:06 PM Page 477
`
`21. E-Trading Systems in Foreign Exchange Markets
`
`477
`
`Table 2
`Interbank Trading Optionsa
`
`Incoming (nonaggressor)
`
`Outgoing (aggressor)
`
`Direct
`
`Indirect
`
`Give quotes on request (market
`making)
`Dealer gives quote(s) to a broker
`(limit order)
`
`Trade at other dealers’ quotes
`(“taker”)
`Dealer trades at quotes given by a
`broker (market order)
`
`aA market maker gives quotes (buy and sell prices) on request from other dealers. The dealer
`that takes the initiative and asks for quotes is called the “aggressor.” Direct trading is bilat-
`eral trade over the D2000-1 computer system or the telephone. Indirect trading is trading
`through a broker, either a traditional “voice” broker (closed telephone network) or the elec-
`tronic brokers D2000-2 and EBS.
`
`makers’ quotes (outgoing trade). Because this is bilateral, it is often called
`direct trading. Furthermore, the interbank market is a hybrid market in the
`sense that liquidity can be provided both through making markets and by
`entering limit orders with brokers. The brokers announce the best bid and
`ask prices, and trading on these is a market order. Brokered trades are often
`called indirect trading.
`The information signal in trading is connected to the action of the most
`active part in the trade, often called the initiator or aggressor (outgoing
`trade). If the aggressor buys (sells), we say that it is a positive (negative)
`order flow, so order flow is just a transaction with a sign. How is this infor-
`mative? Think of the portfolio shift model of Evans and Lyons mentioned
`in the discussion of Table 1. If the aggressor buys, that could be because she
`is covering a position after a customer purchase (portfolio shift into that
`currency). Alternatively, think of the proposition that order flow reveals
`information about other dealers’ expectations. Then a purchase on behalf
`of the aggressor could be a signal, with noise, that the aggressor believes
`the currency is undervalued. In both cases, the positive order flow signals
`that the exchange rate should appreciate.
`
`D. TRANSPARENCY
`
`As mentioned earlier, transparency is low in foreign exchange markets
`compared to most equity markets. There are many forms of transparency:
`pretrade and posttrade transparency, transparency of prices or trades, and
`transparency with respect to whether the customers or only the dealers can
`observe the trading process. To start with the last of these, in the foreign
`
`

`

`EWE21 7/24/03 9:06 PM Page 478
`
`478
`
`Rime
`
`exchange markets only the dealers can observe anything other than their
`own trading. The trades with customers that initiate the trading sequence
`are only observed by the bank that receives the order and, hence, are
`private information for the banks. In the interbank market, trades that are
`made directly between two dealers are only observed by the two dealers.
`The lack of disclosure requirements also ensures that these trades will not
`be observed by other dealers after the trade is made. Indirect trading is
`more transparent because the price and sign (buying or selling by the
`aggressor) of the last trades are observable. In this sense, trading through
`brokers is what determines the level of trade transparency. This level has
`evolved endogenously as a result of dealers’ indirect trading.
`Price transparency is higher than trade transparency, but until recently
`customers’ ability to see prices was less than that of dealers. However, com-
`pared to many equity markets with indirect trading, transparency is still low.
`In many equity markets a trader may be able to see the identity of the best
`bid and ask and often also a part of or the entire order book (all the other
`limit orders). In the case of a trade, both the size and the identity of the
`counterparties are revealed in many equity markets.
`
`III. DIRECT TRADING AND VOICE BROKERS
`
`In this section, we provide a further elaboration on the working of the
`interbank market prior to the electronic brokers by discussing how indirect
`and direct trading actually functions and affects dealers’ behavior. As men-
`tioned earlier, the trading institutions have been more or less unaltered for
`a long time, perhaps since the early 1930s when the first telephone brokers
`started. The composition of direct versus indirect trading has changed over
`time, however. For telephone brokers, the main innovation came in the
`1960s when brokers started operating through private telephone networks.
`These are installed free of charge in banks by brokers. The broker
`announces the best (limit order) bid and ask prices over intercoms at the
`dealers’ desk. If the dealer wants to trade at a limit order, i.e., submit a
`market order, he picks up the phone with the direct line and just says
`“mine” if he is buying (at the ask price) and “yours” if he is selling (at the
`bid price). The voice broker then knows which of the two announced prices
`at which he is trading. After a trade the broker announces the price and
`whether it was traded on the bid or the ask price. The size of the trade is
`not announced, but standard sizes are 1 and 5 million. This announcement
`was the only signal on marketwide order flow that the dealer received. Tele-
`phone brokers are often called voice brokers due to the announcements
`over intercom systems. Voice brokers were very popular up to the mid-
`
`

`

`EWE21 7/24/03 9:06 PM Page 479
`
`21. E-Trading Systems in Foreign Exchange Markets
`
`479
`
`1980s at least. The main advantage for the dealers is that the dealer stays
`anonymous until the trade is made.
`Direct trading was made by telephone or telex in the 1970s. In February
`1981, Reuters introduced the Reuters Market Data Service (RMDS), which
`was like a bilateral bulletin board for conveying trading interest, for sub-
`sequent trading over the telephone. This system was replaced in 1987 by
`Reuters Dealing 2000-1, a closed network for bilateral electronic commu-
`nication. Although a system for electronic trading, it did not revolutionize
`the market. The D2000-1 is more like an advanced telephone and made the
`direct trading that used to take place over the telephone more efficient.
`D2000-1 quickly became the dominant tool for trading bilaterally. The
`dealers “chat” in much the same manner as with “instant messengers” on
`the Internet. Furthermore, trade tickets, needed to check trades and for set-
`tlement with counterparties, were sent automatically to the back office, and
`the dealer could trade faster and more efficiently with up to four conver-
`sations going simultaneously. With this system in place, direct trading
`started to take market share from voice brokers, and in the late 1980s to
`early 1990s, interbank volume was split 50/50 between direct trading and
`voice brokers.
`Table 3 reports some volume numbers from the United Kingdom, United
`States, and Japan, the three largest single markets, to help us get an idea of
`the size of the market. First we notice that in London alone there is trading
`for over $500 billion each day, down from over $630 in 1998. The foreign
`exchange market had grown rapidly from the collapse of Bretton Woods in
`the early 1970s, until the downturn in volume that we see from 1998. Total
`volume has decreased similarly, down to $1200 billion in 2001 from $1490
`billion in 1998. We will come back to the downturn in the next section. The
`increase in volume through the 1980s and 1990s was primarily driven by
`increased globalization, a dramatic increase in trading with customers in the
`late 1980s, and more banks entering the foreign exchange market.The intro-
`duction of D2000-1 might, however, have been a useful trading tool in this
`process. Not only could the system handle more trades simultaneously (the
`dealer could contact four market makers simultaneously) but D2000-1 also
`made cross-border trading easier. Voice brokers are quite regional. There
`are New York-based, London-based, Frankfurt-based, and Tokyo-based
`brokers, brokers serving Scandinavia, etc. Chatting electronically seems to
`be less hampered by borders.
`Notice also the high interbank share of foreign exchange volume in Table
`3. During the 1990s, the interbank share was between 60% and 80%,
`possibly at the high end for financial centers. This has been interpreted as
`speculative trading on the part of the banks, because it cannot be related
`to goods trading, etc. However, within this trading structure with a high
`
`

`

`EWE21 7/24/03 9:06 PM Page 480
`
`480
`
`Rime
`
`Table 3
`Daily Trading Volumesa
`
`United Kingdom
`
`United States
`
`Japan
`
`Volume Share (%) Volume Share (%) Volume Share (%)
`
`2001 Total volume
`Total interbank
`Total spot
`
`1998 Total volume
`Total interbank
`Total spot
`
`1995 Total volume
`Total interbank
`Total spot
`
`1992 Total volume
`Total interbank
`Total spot
`
`1989 Total volume
`Total interbank
`Total spot
`
`1986 Total volume
`Total interbank
`Total spot
`
`504
`341
`151
`
`637
`530
`217
`
`464
`349
`186
`
`290
`230
`148
`
`184
`161
`119
`
`90
`82
`66
`
`68
`30
`
`83
`34
`
`75
`40
`
`79
`51
`
`88
`65
`
`91
`55
`
`254
`143
`104
`
`351
`171
`148
`
`244
`138
`134
`
`167
`122
`95
`
`125
`116
`81
`
`59
`54
`34
`
`56
`41
`
`49
`42
`
`57
`55
`
`73
`57
`
`93
`65
`
`93
`42
`
`147
`123
`37
`
`136
`109
`57
`
`161
`118
`55
`
`120
`92
`48
`
`111
`78
`46
`
`48
`32
`19
`
`83
`25
`
`80
`42
`
`74
`34
`
`76
`40
`
`70
`41
`
`67
`41
`
`aSource: Average daily volumes as reported in BIS (1990, 1993, 1996, 1999, 2002). All volumes
`are corrected for local double-counting. Total volume and total interbank volume in each
`country is the sum of spot, forward, and swap volumes in the overall and interbank markets,
`respectively. Interbank share and spot share are calculated as the share of total interbank
`volume.
`
`share of direct trading, a high interbank share may be the result of risk
`sharing between the banks after receiving large customer orders. Imagine
`that a large customer order, for example €100 million, ends up at the desk
`of an interbank dealer. Let us assume that the dealer wants to get rid of it.
`The dealer has relationships with 10 other dealers (market makers) and
`sells €10 million to each through (outgoing) direct trading. For the sake of
`the argument, assume that none of these dealers are particularly interested
`in the position. They accept the trade because they get compensation
`through the bid–ask spread (incoming trade on their behalf). Each of them
`turns to two of their contacts and sells €5 million to each. The interbank
`volume is now €200 million, and it continues to grow. The €5 million that
`20 dealers have received are sold to other dealers again, and the volume
`
`

`

`EWE21 7/24/03 9:06 PM Page 481
`
`21. E-Trading Systems in Foreign Exchange Markets
`
`481
`
`reaches €300 million! The customer trade is passed on like a “hot potato.”
`When the process comes to an end, all dealers, including our initial dealer,
`hold a share of the initial customer order. Large interbank trading flows
`could, in other words, be a consequence of a market structure with a high
`share of direct trading.
`
`A. DEALER BEHAVIOR
`
`How do the dealers behave in such a hybrid market? We can use the
`dealer studied by Lyons in 1992 (see Lyons, 2001a) as an example of how
`a market maker works. The dealer operated as a market maker in a New
`York investment bank in 1992 and traded almost entirely by giving quotes
`on the D2000-1 system (market making e.g., direct incoming trading).
`A market maker sets bid and ask prices, the difference being the spread
`and the midpoint typically being her expectation. The spread is a function
`of three components: (i) adverse selection protection; (ii) risk management;
`and (iii) order processing costs and rents. To discourage informed traders
`and make money from the uninformed (she always loses to the better
`informed), the market maker increases the spread with trade size, hence,
`making it more expensive to trade. Similarly, the spread increases with
`size as compensation for taking on the risk in the trade. The part of the
`spread due to order processing costs and rents is usually modeled as a
`constant.
`The spread is measured in “pips,” with one pip being the fourth decimal
`in most exchange rates (the fifth in £ exchange rates). The median spread
`for the dmark/$ dealer studied by Lyons (2001a) was three pips, and the
`median trade size was $3 million. Geir Bjønnes and Dagfinn Rime find, in
`similar data for direct trading from 1998, a median spread of two pips, with
`a median trade size of $1 million. The spread was relatively constant up to
`$5 million. This may seem like a tiny transaction cost. If the dmark/$ was
`trading at 1.8, then a two-pip spread is only slightly more than one basis
`point (1%/100), and buying $1 million would cost approximately $55. When
`one realizes that, in April 1998, dmark/$ was traded for almost $100 billion
`daily in the interbank market alone (corrected for double counting), the
`risk sharing process mentioned earlier becomes quite expensive ($5.5
`million daily in interbank dmark/$ trading alone).
`The dealer that makes the contact (aggressor) asks for bid and ask prices
`for a given size without revealing his trading intentions. In “direct” trading,
`market makers are expected to give tight quotes promptly on request, and
`the aggressor is similarly expected to reply quickly. The quotes are on a
`take-it-or-leave-it basis. If there is a trade, the server analyzes the conver-
`
`

`

`EWE21 7/24/03 9:06 PM Page 482