«Volume Title: The Microstructure of Foreign Exchange Markets Volume Author/Editor: Jeffrey A. Frankel, Giampaolo Galli, Alberto Giovannini, editors ...»
This PDF is a selection from an out-of-print volume from the National
Bureau of Economic Research
Volume Title: The Microstructure of Foreign Exchange Markets
Volume Author/Editor: Jeffrey A. Frankel, Giampaolo Galli, Alberto
Volume Publisher: University of Chicago Press
Volume ISBN: 0-226-26000-3
Volume URL: http://www.nber.org/books/fran96-1
Conference Date: July 1-2, 1994
Publication Date: January 1996
Chapter Title: Bid-Ask Spreads in Foreign Exchange Markets: Implications for Models of Asymmetric Information Chapter Author: David A. Hsieh, Allan W. Kleidon Chapter URL: http://www.nber.org/chapters/c11362 Chapter pages in book: (p. 41 - 72) Bid-Ask Spreads in Foreign Exchange Markets: Implications for Models of Asymmetric Information David A. Hsieh and Allan W. Kleidon The term market microstructure was coined in 1976 by Mark Garman to define "moment-to-moment trading activities in asset markets." With the stated goal of providing insight and testable implications regarding the transaction-totransaction nature of realistic exchange processes, Garman examines dealership and auction models of marketmakers where the stream of market orders from a collection of market agents is depicted as a stochastic Poisson process.
The resulting insights concern bid-ask spreads (based on standard microeconomic analysis), inventories of marketmakers, and the effects of some market power on the part of marketmakers.
A more recent microstructure literature is based on information asymmetries among economic agents. In a recent literature review, Admati (1991) defines the area of market microstructure as "the literature on asset markets with asymmetric information and especially on trading mechanisms" (p. 347); Garman is noted as an example of "earlier market microstructure literature" in which "information issues were not typically modelled" (p. 355, n. 11). Two reasons are given for the popularity of asymmetric information models: policy implications of different trading processes, as exemplified by the 1987 crash, and empirical results such as various patterns in trading volume, variances, and returns over the trading day. The belief is that better insights into both issues will be achieved by examining information asymmetries. Moreover, observed empiriDavid A. Hsieh is professor of finance and economics at the Fuqua School of Business of Duke University. Allan W. Kleidon is vice president at Cornerstone Research and a consulting professor of law (in finance) at the School of Law of Stanford University.
Support for this research has been provided by Cornerstone Research. The authors are grateful for helpful comments from Jeffrey Frankel, Paul Pfleiderer, Antti Suvanto, currency traders at Citibank and at Wells Fargo-Nikko Investment Advisors (Jeffrey Hord and Vikas Srivastava), two anonymous referees, seminar participants at the Haas School of Business, University of California, Berkeley, and participants at the conference on the microstructure of foreign exchange markets.
42 David A. Hsieh and Allan W. Kleidon cal results are believed to "call for theoretical explanations beyond what can be obtained by traditional models (in which informational asymmetries are not present)" (p. 348).
In some ways, foreign exchange data have institutional features that are ideal for testing these now "standard" asymmetric information models. The market is very liquid and is linked around the world by computerized information systems, and the commodity is essentially the same in all markets. Standard information models have been applied to foreign exchange data in Bollerslev and Domowitz (1993), particularly to the behavior of bid-ask spreads and volume around the open and close of trading in regional markets. Bollerslev and Domowitz conclude that data for smaller banks that "operate mainly within regional markets with well-defined market openings and closings" (p. 1422) show the relation between trading activity and spreads that is implied by standard asymmetric information models.
In this paper, we further examine how well standard asymmetric information models can explain the behavior of volatility, bid-ask spreads, and volume around the open and close of trading in foreign exchange markets.1 We conclude that the standard information models are unable to explain these data.
Our analysis differs from most previous studies in that we examine the implications of standard information models for the behavior of data across markets that are open simultaneously, rather than looking at markets in isolation.
In particular, a feature of standard information models is that high volatility is associated with trades by privately informed traders whose trading activity incorporates their information into prices and quotes. Within this class of model, if new information results in high volatility of quotes for a trader located in London, then the quotes for a trader who is physically located in New York but who observes the London quotes will also show high volatility.
We exploit the fact that foreign exchange transactions occur virtually around the clock, with overlap between the trading day for traders in London and New York. Consequently, the open of trade in New York and the close of trade in London correspond to times when the other market has been trading for some time. We find that the high volatility that shows up at the open in New York and the close in London appears to be unrelated to the concurrent volatility in the other market, even though both sets of quotes appear on exactly the same trading screens at exactly the same time.
This volatility cannot be due to new information reaching one market but not the other, within the standard information models. Either these markets that are ostensibly closely linked are segmented in important ways not recognized in standard models, or some phenomenon other than the incorporation
1. Trading patterns at the open and close of trade have been extensively studied within asymmetric-information models, primarily with respect to the New York Stock Exchange (NYSE), but also in the context of cross-country listings and foreign exchange (see, e.g., Admati and Pfleiderer 1988; Subrahmanyam 1991; Freedman 1989; Barclay, Litzenberger, and Warner 1990;
and Bollerslev and Domowitz 1993). For an excellent survey of the literature, see Admati (1991).
43 Bid-Ask Spreads in Foreign Exchange Markets of private information must be responsible for the behavior of quotes. Given the high degree of integration of the international foreign exchange market, we conclude that the observed periodicity of volatility is not due to the incorporation of private information as envisioned by standard asymmetric information models. One way of stating the problem is that if no new information is reaching the international foreign exchange market—which is implied by the absence of unusual volatility in quotes generated by traders in one physical location—then quotes generated by traders in another market show excess volatility relative to that implied by standard information models.
This is not a new phenomenon. For example, the crash of October 1987 is an example of a large change in stock prices that does not appear to have been caused by new information reaching the market as a whole. However, a recent class of model has been developed that explains such phenomena as the crash in terms of imperfect information aggregation and learning by market participants rather than new information reaching the market as a whole. These models differ from standard asymmetric information models by relaxing the assumption that each trader has perfect knowledge about the structure of the market, that is, about the preferences and beliefs of all traders in the market.2 While this is a different type of model from standard asymmetric information models, it is possible that at least some of the observed behavior in foreign exchange markets may be attributable to this type of information asymmetry.
We examine this possibility below, after closely examining the standard asymmetric information models. It appears that some form of learning about the market structure is important at the start of trading, which results in wide and volatile quotes when traders first enter the market. This process may be as informal as the posting of wide quotes with little expectation of trading during the initial period of learning.
At the close of trading, standard information models are again unable to explain the foreign exchange data. We conclude that inventory management by marketmakers in the closing market appears to be the most likely explanation.
The paper proceeds as follows. Section 2.1 discusses the institutional features of our data and presents our empirical results. Section 2.2 examines the standard asymmetric information models of intraday price and volume of Admati and Pfleiderer (1988) and Subrahmanyam (1989, 1991) and the current application of these models to foreign exchange data in Bollerslev and Domowitz (1993).3 Section 2.3 examines alternate explanations of the results in
2. These issues are addressed in Gennotte and Leland (1990), Jacklin, Kleidon, and Pfleiderer (1992), Kleidon (1992, 1995), and Romer (1993).
3. Foster and Viswanathan (1990) present a related asymmetric information model that focuses on interday rather than intraday variations in price and volume. Although it can be argued that foreign exchange trading is not tied to any particular "day," our current focus is on behavior between the open and the close of usual trading hours for traders within the geographic markets of London and New York. Foster and Viswanathan examine the behavior of trading across days, particularly the effects of weekends on Monday trading.
44 David A. Hsieh and Allan W. Kleidon section 2.1, namely, recent learning models based on asymmetric information and other market microstructure models that place more emphasis on the roles of market power and inventory management of marketmakers. Section 2.4 contains concluding remarks.
2.1 Data and Results This section examines the behavior of spreads and quote volatilities across different regional foreign exchange markets at the same point in time. We demonstrate that a key implication of standard asymmetric information models is rejected in foreign exchange data, namely, that periods of high variance correspond to periods of high concentration of informed trading. When one regional market has high variance (i.e., open and close of the regional market with concurrent high bid-ask spreads), other markets simultaneously have low variance (and low bid-ask spreads), even though the traders from different markets are connected by computer terminals with all quotes appearing simultaneously on all terminals. In section 2.2, we interpret these results as showing that whatever the explanations for these phenomena—and we suggest possibilities in section 2.3 below—they are not consistent with current standard models of asymmetric information.
2.1.1 Data The foreign exchange market can be roughly divided into two groups.4 The first group comprises marketmakers or the interbank market, which accounts for most foreign exchange trading.5 Marketmakers deal with each other through a very active computerized market that trades virtually around the clock, either directly or through interdealer brokers. The second group comprises the retail market or customers who approach a local broker or bank and are offered retail foreign exchange quotes by that retail bank.
The interbank foreign exchange market, from which our data are obtained, comprises a network of major trading banks throughout the world that are linked interactively via computer screens (either Reuters or Telerate systems).
We use data from the Reuters indications system which transmits computerized quotes among interbank dealers. When a trading bank individually updates its quotes, the new quotes directly appear on the screens of all traders around the world. Actual trades are consummated via telephone,6 and price and
4. For an excellent description of the markets, see Burnham (1989). See also Goodhart (1990), Goodhart and Figliuoli (1991), Lyons (1992, 1993, 1995), and Bollerslev and Domowitz (1993).
5. Lyons (1993, 2), citing the New York Federal Reserve Bank, states that over 80 percent of trading volume is between marketmakers.
6. Lyons (1995) examines data for a single marketmaker from the Reuters Dealing 2000 System that allows screen trading, although direct telephone communication was necessary for all traders when our data were collected. Since our approach requires cross-market comparisons, the new data set is insufficient for our purposes. It will be useful to replicate our study using data from the new Reuters system, if sufficient data ultimately become available.
45 Bid-Ask Spreads in Foreign Exchange Markets volume for direct interdealer trades are not publicly revealed. Some information about brokered interdealer trades, namely, price, quantity, and whether the trade is at the bid or the ask, is publicly disseminated to dealers via an intercom system. Major trading banks often perform both interbank and retail roles, with a dedicated foreign exchange desk within the bank linked to the interbank market and with retail customers who are offered quotes that consist of the dealers' interbank quotes plus an additional markup.
The deutsche mark/dollar data that we use were originally captured from a Reuters data feed by Charles Goodhart and cover the eighty-two days from 9 April to 30 June 1989. These are the same data used by Bollerslev and Domowitz (1993), who provide valuable descriptions of the characteristics of these data.7 For our purposes, we concentrate on two markets—London and New York—but our analysis applies to the other markets documented in Bollerslev and Domowitz. Figures 2.1a and 2.1b document the time periods in which significant trading activity takes place in London and New York, respectively.