مقاله سال ۲۰۱۷

Globally, the foreign exchange (FX) market is the largest financial market, with an average of approximately $5.3 trillion transactions per day (BIS, 2013). This large market has special characteristics. First, the market is composed of a handful of participants. In interdealer FX markets, trading is concentrated among a few of large financial institutions (King et al., 2013). Second, unlike a stock market, trading is not centralized in FX markets. Although the recent spread of the electronic broking system has centralized trading to some extent, no unique system is used to trade a certain currency pair, and decentralized transactions for one currency pair occur through different systems. Additionally, BIS (2010) reports that approximately 60% of trades still occur through non-electronic broking systems, and most of these trades are over-the-counter (OTC). A handful of participants engage in decentralized trading in the FX market, and Menkhoff et al. (2013) call this opaque market a “dark” one.

Does this dark market achieve informational efficiency? Compared with a stock market, is this market more (less) efficient? As reported in studies on stock markets (e.g., Chordia et al., 2005, 2008; Chung and Hrazdil, 2010), is a market’s efficiency related to its conditions, such as liquidity? These questions motivate this research on the EUR/USD and the USD/JPY markets.

To examine these questions, I propose measures of market efficiency in an FX market. The proposed measures contribute to the microstructure analysis in the following respects. First, these measures are available at high frequencies and help microstructure researchers. For example, when one considers one-minute order flow predictability, these proposed measures are available at every one-minute interval, allowing researchers to engage in high frequency analysis (e.g., Tables 3 and 5 in this study). Second, these proposed measures seem economically more relevant than previous ones when one considers the predictability of order flows given that one of the proposed measures (EF2) considers tangible gains from the order flow signal. Third, the proposed measures are simple for calculations and easily applied to research on other markets in which order flow information is available. Fourth, FX market efficiency has been discussed primarily in terms of uncovered interest rate parity (UIP) holding (see Sarno and Taylor (2002)), and I expect that the proposed measures provide additional insights into an analysis of FX market efficiency.