«Volume Title: The Microstructure of Foreign Exchange Markets Volume Author/Editor: Jeffrey A. Frankel, Giampaolo Galli, Alberto Giovannini, editors ...»
In this paper, Hsieh and Kleidon begin with a precise description of the state of the "market microstructure literature." They remind the readers of the original definition of market microstructure by Mark Garman (1976) and summarize the current dominating paradigm in the literature—the asymmetric information models in the line of Admati (1991), Admati and Pfleiderer (1988), and others. This introduction is useful to exchange rate researchers attempting to exploit microstructure models. It serves as a warning that, given its narrow focus, we should not expect this literature to give us answers to all the macroeconomic puzzles related to foreign exchange rates. On the bright side, this paper has made a welcome attempt to expand the current literature with alternative paradigms. A broader intellectual base may well be what we need to understand the behavior of the foreign exchange market.
The empirical evidence amassed in the paper clearly indicates that the opening and closing of New York and London markets seem to be local events. The associated trading activity and volatility do not transmit from one location to the other, even though the two markets are open to trade at the same time.
This evidence could be interpreted in two different ways. First, the asymmetric information model may be wrong, in the sense that private information associated with high volatility in one market is not picked up by the other market.
Given the simultaneity of information processing on the Reuters screen in the two markets, and given that currencies are the most standardized vehicle of trade, this interpretation does not stand well. This leads to the second interpreZhaohui Chen is the Jean Monnet Lecturer in the Economics of the European Union at the London School of Economics and a research affiliate of the Centre for Economic Policy Research.
66 David A. Hsieh and Allan W. Kleidon tation: the unusual market activities around the opening and closing of the markets are not driven by private information. Under this interpretation, the evidence should not be seen as a rejection of the asymmetric information model since the events used are not information driven and therefore cannot be used to test the information transmission story behind the Admati and Pfleiderer-type models. However, the evidence and the institutional analysis in the paper unambiguously show that the conventional asymmetric model is not applicable to the problems of this paper. This is an important message since it proves, by counterexamples, that the currently dominating paradigm in the microstructure literature is inadequate to cover a possibly broad array of issues in the foreign exchange market.
The second half of the paper explores different paradigms, including the traditional inventory models and the new models of learning. At this stage, most of the learning models are based on experimental results. Nevertheless, it would be extremely interesting if the authors, or future researchers, could apply their econometric tools to test the applicability of such theories and to provide evidence that can help validate different assumptions behind the new theories.
Following the discussion of the learning models in the paper, I would like to raise two questions. One is what the traders try to learn in the process; the other is whether there is a welfare gain from learning. Regarding the first question, the paper mentioned specific parameters such as preferences and beliefs and the ambiguous notion of the "feel" of the market. One possible explanation is that traders need to learn each other's trading rule, or asset valuation scheme, and even the theories behind them, as they do not want to be the outliers in the market. They can do that by testing each other's reaction to various quotes, hence the large number of quotes and high volatility from the screen data around the opening of the market. Regarding the second question, my hypothesis is that a pricing rule formed through consultation with other market participants is more likely to be correct (in absolute terms) than a simple average of other people's quotes (the whole is larger than the sum of the parts). This point is illustrated plainly (or mysteriously) in the popular management experimental game known as "Desert Survival Game," where a group of people are asked to form a survival strategy after an imaginary plane crash in a desert. The group decision following a full consultation among the group members is almost always better than the simple average of individual opinions before the consultation.
Finally, the use of the Reuters quote data in this paper is open to question in light of the recent study by Goodhart, Ito, and Payne (chap. 4 in this volume), in which they find no linkage between these quotes and real trading activities as captured by a more up-to-date data screen. Given the dismissal of the private information story, the problem is less damaging to this paper because, as the later part of the paper suggests, there may be different reasons leading to the 67 Bid-Ask Spreads in Foreign Exchange Markets use of quotes, such as learning and testing ("feeling") the market, that are not necessarily linked to actual trading activities.
References Admati, A. R. 1991. The information role of prices: A review essay. Journal of Monetary Economics 28:347-61.
Admati, A. R., and P. Pfleiderer. 1988. A theory of intraday patterns: Volumes and price volatility. Review of Financial Studies 1:3-40.
Garman, M. B. 1976. Market microstructure. Journal of Financial Economics 3:257-75.
Comment Antti Suvanto
The authors make the empirical observation that volatility of the deutsche mark/dollar exchange rate and the spread are high at both the open and the close of the trading day. This U-shaped pattern is evident for both New York and London. Because of time-zone differences, London opens when New York is still closed. When New York opens, London still has a few hours to go.
The second observation made in the paper is that the higher volatility and wider spread in New York at the time of opening are in no way reflected in the volatility and spread in the London market. Similarly, the higher volatility and wider spread at the time London closes do not affect midday volatility and spread in New York.
These observations seem to contradict predictions of recent microstructure theory based on asymmetric information between informed insiders and liquidity traders. This theory has been used to explain the similar U-shaped intraday pattern of the volatility and the spread in organized securities markets, such as the New York Stock Exchange.
The authors argue that, because the foreign exchange market is globally integrated and the object of trade the most homogeneous of all, the participation of informed traders at the beginning of the trading day in New York should be visible in London as well. Therefore, the spread should widen and the volatility increase at the same time in both markets. The same should be true if informed trading in London is concentrated toward the end of the trading day. Because this is not the case, the asymmetric information argument is not applicable to the foreign exchange market.
The authors suggest two types of explanations for these observations. The first is based on the heterogeneity of information across marketmakers. The Antti Suvanto is head of the Economics Department of the Bank of Finland.
68 David A. Hsieh and Allan W. Kleidon second is based on the inventory (position) adjustment behavior of "day traders." I find these suggestive explanations plausible, and I agree with the authors' final conclusion that further investigation along these lines is warranted.
Position Constraint Let us take the inventory argument first. The authors point out that a large proportion of foreign exchange trading is "day trading," by which they mean that the banks open and close the day with closed positions. I am not sure whether it is accurate to refer to smaller regional banks alone in this context because all banks, including branches of the big international banks, are "day traders." They stop trading at some point in the day, whether their overnight position is open or not. Otherwise, there would not be distinct but overlapping trading days in London and New York.
The point made by the authors is, however, relevant. The fact that important news may be disclosed during the time when the bank is not trading and the fact that trading continues in other time zones imply that the dealers must pay attention to overnight positions. As the trading day winds down, dealers must either close a position or aim at a desired fixed open position. Whatever the position target for the end of the day, the problem of achieving this target becomes increasingly acute toward the end of the trading day, as the authors correctly point out.
I have shown elsewhere that this is true in a partial equilibrium model of the pricing behavior of an isolated profit-maximizing marketmaker facing stochastic but price-sensitive order flow from uninformed customers and aiming at a closed position or at any other fixed position target for the end of the day (Suvanto 1993, chap. 2).
Applying dynamic programming techniques, the resulting sequential pricing rule implies weak efficiency of the mid-rate and low-volatility of prices (relative to the volatility of order flow) during the most of the day, with volatility increasing toward the end of the day. The spread is chosen so as to maximize the expected trading income at each short trading period given the price sensitivity of customer orders.
The reluctance of the dealer to make frequent and drastic changes in the quoted two-way price in the absence of new information derives from the fact that such changes are generally revenue reducing, that is, selling at a low price a currency just purchased at a high price. Note that in this model the binding two-way price is set before the direction and size of the trade are known. The counterpart to small price adjustments is the dealer's willingness to accept comparatively large open intraday positions. In fact, the ex ante position variance increases at the beginning of the day and starts to decline toward the end of the day.
Although the environmental assumptions of this (monopolist) model are far from those of the highly competitive and closely integrated foreign exchange 69 Bid-Ask Spreads in Foreign Exchange Markets market of today, the results can be generalized to a more realistic setting. Opening the model to competition and interdealer transactions makes the quotation of an individual dealer dependent on the price quotes of others and narrows the spread. From the position-adjustment point of view, the life of an individual dealer is easier because he can undo an unwanted position by calling another marketmaker or he can stay on the "right side of the market" and attract orders from other dealers by making a very small price adjustment.
This is true for the most of the day, but the mechanism does not work in the same way just before the close. Dealers who are already satisfied with their positions do not want new orders. Widening the spread decreases the likelihood of new orders. There may be unwanted open positions that do not find buyers until the spread is sufficiently large to make the price attractive to somebody, or sufficiently unattractive to the seller, until either of them is willing to carry an open overnight position.
Allowing for overlapping time zones eliminates this problem because dealers in a closing time zone can sell unwanted open positions to buyers in open time zones. Dealers in the latter are happy to buy these positions because they are on the "right side of the market" and have plenty of time to undo the position during the remainder of the day.
Heterogeneous Information Let us now turn to the heterogeneity of information. Information may be heterogeneous even if all participants have the same information but interpret its significance for the exchange rate somewhat differently. Economists very often have the same information. Yet they may permanently disagree on the interpretation. Could one reasonably expect dealers always to have a uniform interpretation of the remainder of the day?
The authors point out correctly that dealers need to get the "feel" of the market before engaging in active trading. What does this feel mean?
Assume that at the time of the London opening all dealers have the same information on the macroeconomic and political news that has been reported since the previous close and that they have the same expectations about news that is forthcoming during the day. Assume further that each dealer knows with probability one that all nonbank customers that may appear during the day are price-sensitive liquidity traders. These rather extreme assumptions should eliminate all information asymmetries at the beginning of the day, but this is not the case.
Depending on the situation, each dealer acts in two different roles in relation to his interbank counterparties. Sometimes he is on the "right side" of the market and acts as a marketmaker in relation to another dealer who is requesting a quote. On other occasions, he is on the "wrong side" of the market, requesting a quote from someone else.
Each time dealer A quotes a two-way price to another dealer, B, he does not know exactly what prices are quoted by his rivals. In particular, he does not 70 David A. Hsieh and Allan W. Kleidon know whether at the very same moment his dealer-customer B is receiving from dealer C a price that is inconsistent with his own quote in the sense that his ask price is below the bid price quoted by dealer C (or vice versa). If this is the case, there is an arbitrage opportunity for dealer B, who would never leave such an opportunity unexploited.
I have argued elsewhere that the possibility of inconsistent quotes cannot be eliminated entirely as long as two-way prices are quoted independently without full knowledge of all other quotes at the same moment (Suvanto 1993, chap. 3). This brings potential insiders into the picture because some participants may have observed a pair of inconsistent prices. This explains the positive bid-ask spread in the same way as the possibility of informed traders explains the positive spread in the stock market. Fundamental valuation efficiency plays no role in this context.