یورو به عنوان ارز بین المللی: تبیین اولین شواهد شگفت انگیزتر از بازار ارز خارجی

This paper presents evidence that the bid-ask spreads in euro rates increased relative to the corresponding bid-ask spreads in the German mark (DM) prior to the creation of the currency union. This comes with a decrease in transaction volume in the euro rates relative to the previous DM rates. The starkest example is the DM(euro)/yen rate in which the spread has risen by almost two-thirds while the volume decreased by more than one third. This outcome is surprising because the common currency concentrated market liquidity in fewer external euro rates and higher volume tends to be associated with lower spreads. We propose a microstructure explanation based on a change in the information environment of the FX market. The elimination of many cross currency pairs increased the market transparency for order flow imbalances in the dealership market. It is argued that higher market transparency adversely affects the inventory risk sharing efficiency of the dealership market and induces the observed euro spread increase and transaction volume shortfall.

A key feature of an international currency is a dominant role in the global foreign exchange (FX) market.1 It has long been realized that this role can by far surpass a country’s relative importance in international trade. Examples are the asymmetric role of the dollar in global foreign exchange trading or of the German mark (DM) for European currency pairs. This observation lead to the notion of a “vehicle currency,” which emerges from currency competition as a predominant transaction medium because of network externalities and scale economies translating high transaction volume into lower transaction costs. This vehicle currency perspective is underpinned by evidence on the inverse long-run relationship between transaction costs and transaction volume across different foreign exchange markets (Hartmann, 1998a, Hartmann, 1998b and Hartmann, 1999).2 The creation of the euro implied a liquidity consolidation of many external rates in a single euro rate. The euro could therefore hope to surpass the German mark as a vehicle currency in the global FX market.3 The first part of our paper examines the outcome of the “euro experiment” for the FX market in the first one to two years of the montary union. We examine evidence both on external bid-ask spreads and transaction volumes in the pre- and post-euro environment and find that both criteria indicate a diminished role of the euro relative to the German mark. We can summarize our empirical findings in three points: Stylized Fact 1 (Euro Transaction Costs). The euro is characterized by higher transaction costs in the dollar, yen and sterling markets. Quoted spreads increased from approximately 3.76 to 5.26 basis points for the dollar rate, from 5.1 to 8.3 basis points for the yen rate and from 3.1 to 9.2 basis points for the sterling rate. Other evidence strongly suggests that this widening extends to effective spreads. Stylized Fact 2 (Non-euro Transaction Costs). The increase in the cost of euro transactions, relative to DM transactions, occurred against a background of constant or even declining costs in dollar markets, other than the euro/dollar market. Stylized Fact 3 (Transaction Volume). A unique data set for brokered interdealer trades shows lower euro transaction volumes compared to the previous DM volumes. The decrease in the role of euro transactions, compared to DM transactions, occurred against stable or even increasing transaction volumes in dollar markets, other than the euro/dollar market. To put these results into perspective it is useful to recall previous forecasts about the euro’s transaction role relative to the German mark. Hartmann (1998) derives total volume forcasts for spot trading using data by the Bank for International Settlements (1996). Based on a scenario which excludes the United Kingdom from participation, the statistics in Table 1 were presented as plausible outcomes. The elimination of intra-EU volume (Row (2)) reduces the overall euro volume (Row (4)) relative to the DM volume (Row (1)) and is not fully compensated by the consolidation of external liquidity (Row (3)) in one common currency. However, our own evidence suggests that in reality the external euro volume itself decreased relative to the external volume of the DM in spite of the volume boost from external consolidation. This is a rather surprising result. The second part our paper interprets these puzzling findings and develops a stylized model consistent with the evidence. We argue that the evidence requires a refinement of the existing vehicle currency perspective for currency unions. The positive effect of liquidity consolidation and the long-run inverse volume-spread relationship may well be at work. But this conventional linkage documented for existing currency pairs may have been dominated by a second effect specific to the creation of a currency union. The currency union reduced the number of currencies and therefore changed the information environment of the foreign exchange market. The consolidation of liquidity in a single euro rate increased “market transparency.” By market transparency we mean the knowledge of an individual dealer about the positions and trading desires of all other market participants. Higher market transparency in turn may reduce interdealer risk sharing and increase inventory risk for the dealers. We demonstrate that the equilibrium outcome may be higher spreads and lower volume. The role of market transparency in FX trading has to be appreciated in the light of two key market characteristics. First, the FX market is characterized by low market transparency. The dealership structure limits information revelation on transaction prices to own party trades. Moreover, there exists a wide range of asset substitutes which makes if difficult to discern position and trading desires of other market participants. A short spot position in DM and a long spot position in dollars can be hedged in different instruments (like forwards and futures) or through different cross rates, for example buying dollars for French francs (FF) and simultaneously buying FF for DM. Second, inventory risk management concerns in the FX market are more important than in other dealership markets. The average time span for the inventory cycles of a foreign exchange trader may be as short as 15 minutes. Inventory risk is therefore a crucial determinant of trader behavior and FX spread quotation (Bollerslev and Melvin, 1994; Lyons, 1995 and Lyons, 1996). The creation of the euro implies that all internal cross rates disappear, while external markets are consolidated in a single rate. The reduced number of trading venues is likely to accelerate the information revelation process and increases public information about excess balances held by other dealers. But an important benefit of a dealership market is interdealer inventory risk sharing after large customer orders. The high interdealer trading volumes indicate this important market function. Accelerated information revelation may be harmful to the dealers as it tends to undermine inventory risk sharing. A low transparency environment allows dealers to pass excess balances to other dealers prior to information revelation and price change. A high transparency environment on the other hand may imply a price change prior to full risk sharing. In the latter case the dealer takes a big concentrated loss on his excess balances which he did not manage to diffuse to other uninformed dealers. He can only protect himself against the increased inventory risk by quoting larger spreads. The higher market transparency induced by the currency union may therefore explain the evidence on higher spreads for the euro (Stylized Fact 1). But higher spreads decrease the role of the euro as a vehicle currency without adverse effect for the dollar rates (Stylized Fact 2). The lower transaction volumes in the euro rates follows from the conventional inverse relationship between transaction costs and volume (Stylized Fact 3). The role of changing market transparency under a currency union is difficult to formalize. We provide a very stylized model representation which extends the Copeland and Galai (1983) model to a system of three currencies. This allows us to examine the case of a merger of two currencies in a union and compare the prior market characteristics to those after the union formation. The model is designed to illustrate the benefit of interdealer risk sharing in a system of multiple currencies. We show under which conditions a dealer faced with inventory imbalances from client order flow uses cross hedging opportunities in the adjunct currency pairs to manage his inventory risk. A long position in dollars and a short position in DM can for example be hedged by simultaneously selling French francs for dollars and buying French francs for DM. The maintained model assumption about market transparency is that inventory sharing in the same currency pair is possible only at the full information price, while adjunct markets provide the risk sharing options prior to complete price adjustment. This captures the admittedly extreme case in which dealers have full information about excess balances in one currency pair (in which they may specialize) and none in the adjunct markets. The more general message is that differential information about excess balances in markets with substitutable assets will naturally foster risk sharing. If these risk sharing opportunities decrease or disappear or in the currency union, we obtain a market with higher spreads and possibly lower volume (in spite of the liquidity consolidation) as observed for the external euro currency pairs. The role of market transparency for the spread determination is a very recent concern for the design of equity markets. Evidence that higher market transparency can increase spreads has recently been provided by Bloomfield and O’Hara (1999) in an experimental setting.4 It is interesting to note that some stock exchanges try to offer a choice to market participants about the desired degree of market transparency. The limit order books of the Paris and Toronto stock exchange for example allows traders to post a part of their limit order in the form of a hidden order which does not appear on the otherwise public limit order book. Harris (1996) shows that hiding trading desires becomes more prevalent as the trade size increases. But his evidence also shows that higher transaction costs (measured by minimum tick size relative to asset price) reduce a dealer’s desire to hide orders. This suggests that the inventory costs of increased market transparency may therefore be particularly high in the FX market in which a low spread (compared to equity markets) encourages frontrunning. Madhavan et al. (2000) study the effects of greater transparency in the Toronto Stock Exchange (TSE) before and after the limit order book was publicly disseminated. This natural experiment allows them to isolate the effects of transparency while controlling for stock-specific factors and for type (floor or automated) of trading system. They find that transparency had detrimental effects on execution costs and liquidity. Our paper is organized as follows. The empirical Section 2 starts with a discussion of the foreign exchange market structure and the data sources. Section 2.3 describes the structural break in transaction costs in the external euro rates relative to the previous corresponding DM rates. Section 2.4 uses unique data on brokered inter-dealer transactions to highlight the reduced euro turnover. Section 3 provides a theoretical microstructure framework for interpreting this evidence. Section 4 summarizes our conclusions.

The advent of the euro raised the expectation that this new currency would establish itself as a currency of greater global importance than its predecessor, the DM. Such expectations were grounded in the belief that a currency union is primarily a consolidation of the financial liquidity of the union members and that this liquidity consolidation would be large enough to set in motion substantial network externalities. Our evidence suggests that the contrary took place at the start of EMU, namely euro volumes decreased and euro spreads increased. This suggests that an important theory element was absent in the previous analysis. The scale economy hypothesis underlying the evidence for existing currency pairs and the vehicle currency theory itself may not be in question. But we argue that a currency union changes the information environment in which the new external currency rates of the union are determined. Parallel to the consolidation of liqudity occurs an increase of market transparency as many opaque cross hedging opportunities disappear. This can have important microstructure consequences for the risk sharing efficiency of the dealership market. Under accelerated information revelation, inventory risk increases and pushes up the competitive equilibrium spreads relative to the corresponding rates in the legacy currencies. The market transparency effect can explain why we find that euro transactions are more expensive than former DM transactions. The increase in the spread is associated with a decrease in volume according to the conventional spread-volume linkage which can more than offset the volume enhancing liquidity consolidation. This can explain the low euro rate transaction volumes relative to the DM volumes.