Order Flow and Exchange Rate Dynamics Martin D. D. Evans∗ Richard K. Lyons This draft: December 1999 Abstract Macroeconomic models of nominal exchange rates perform poorly. In sample, R2 statistics as high as 10 percent are rare. Out of sample, these models are typi-cally out-forecast by a naïve random walk. This paper presents a model of a new kind. Instead of relying exclusively on macroeconomic determinants, the model includes a determinant from the field of microstructure—order flow. Order flow is the proximate determinant of price in all microstructure models. This is a radically different approach to exchange rate determination. It is also strikingly successful in accounting for realized rates. Our model of daily exchange-rate changes produces R2 statistics above 50 percent. Out of sample, our model pro-duces significantly better short-horizon forecasts than a random walk. For the DM/$ spot market as a whole, we find that $1 billion of net dollar purchases in-creases the DM price of a dollar by about 0.5 percent. Correspondence Richard K. Lyons Haas School of Business, UC Berkeley Berkeley, CA 94720-1900 Tel: 510-642-1059, Fax: 510-643-1420 [email protected] www.haas.berkeley.edu/~lyons ∗ Respective affiliations are Georgetown University and NBER, and UC Berkeley and NBER. We thank the following for valuable comments: Menzie Chinn, Peter DeMarzo, Frank Diebold, Petra Geraats, Richard Meese, Michael Melvin, Peter Reiss, Andrew Rose, Mark Taranto, Ingrid Werner, Alwyn Young, and seminar participants at Chicago, Wharton, Columbia, MIT, Iowa, Houston, Stanford, UC Berkeley, the 1999 NBER Summer Institute (IFM), and the December 1999 NBER program meeting in Microstructure. Lyons thanks the National Science Foundation for financial assistance.1Order Flow and Exchange Rate Dynamics Omitted variables is another possible explanation for the lack of explanatory power in asset market models. However, empirical researchers have shown considerable imagina-tion in their specification searches, so it is not easy to think of variables that have es-caped consideration in an exchange rate equation. Richard Meese (1990) 1. Motivation: Microstructure Meets Exchange Rate Economics Since the landmark papers of Meese and Rogoff (1983a, 1983b), exchange rate economics has been in crisis. It is in crisis in the sense that current macroeco-nomic approaches to exchange rates are empirical failures: the proportion of monthly exchange rate changes that current models can explain is essentially zero. In their survey, Frankel and Rose (1995) write “the Meese and Rogoff analysis at short horizons has never been convincingly overturned or explained. It continues to exert a pessimistic effect on the field of empirical exchange rate modeling in particu-lar and international finance in general.” 1 Which direction to turn is not obvious. Flood and Rose (1995), for example, are “driven to the conclusion that the most critical determinants of exchange rate volatility are not macroeconomic.” If determinants are not macro fundamentals like interest rates, money supplies, and trade balances, then what are they? Two alternatives have attracted attention. The first is that exchange rate determinants include extraneous variables. These extraneous variables are typically modeled as rational speculative bubbles (Blanchard 1979, Dornbusch 1982, Meese 1986, and Evans 1986, among others). Though the jury is still out, Flood and Hodrick (1990) conclude that the bubble alternative remains unconvincing. A second alternative to macro fundamentals is irrationality. For example, exchange rates may be deter-mined in part from avoidable expectational errors (Dominguez 1986, Frankel and Froot 1987, and Hau 1998, among others). On a priori grounds, many economists find this second alternative unappealing. Even if one is sympathetic to the presence 1 The relevant literature is vast. Recent surveys include Frankel and Rose (1995), Isard (1995), and Taylor (1995).2of irrationality, there is a wide gulf between its presence and accounting for ex-change rates empirically. Until it can produce an empirical account, this too will remain an unconvincing alternative. Our paper moves in a new direction: the microeconomics of asset pricing. This direction makes available a rich set of models from the field of microstructure finance. These models are largely new to exchange rate economics, and in this sense they provide a fresh approach. For example, microstructure models direct attention to new variables, variables that have “escaped the consideration” of macroecono-mists (borrowing from the opening quote). The most important of these variables is order flow.2 Order flow is the proximate determinant of price in all microstructure models. (That order flow determines price is therefore robust to differences in market structure, which makes this property more general than it might seem.) Our analysis draws heavily on this causal link from order flow to price. One level deeper, microstructure models also provide discipline for thinking about how order flow itself is determined. Information is key here—in particular, information that currency markets need to aggregate. This can include traditional macro fundamentals, but is not limited to them. In sum, our microeconomic approach provides a new type of alternative to the traditional macro approach, one that does not rely on extraneous information or irrationality.3 Turning to the data, we find that order flow does indeed matter for exchange-rate determination. By “matter” we mean that order flow explains most of the variation in nominal exchange rates over periods as long as four months. The graphs below provide a convenient summary of this explanatory power. The solid lines are the spot rates of the DM and Yen against the Dollar over our four-month sample (May 1 to August 31, 1996). The dashed lines are marketwide order-flow for the respective currencies. Order flow, denoted by x, is the sum over time of signed trades between foreign exchange dealers worldwide.4 2 Order flow is a measure of buying/selling pressure. It is the net of buyer-initiated orders and seller-initiated orders. In a dealer market such as spot foreign exchange, it is the dealers who absorb this order flow, and they are compensated for doing so. (In an auction market, limit orders absorb the flow of market orders.) 3 Another alternative to