Microstructure Models of Foreign Exchange Markets

Transcription

1 Microstructure Models of Foreign Exchange Markets Tyler Kustra January 28,

2 The advent of market microstructure models, which use trading data to predict exchange rates, stemmed from the failure of models which use macroeconomic data to predict exchange rates with an acceptable degree of accuracy. 1 Macroeconomic models which use variables such as interest rates and inflation have an R 2 between 1 and 5 per cent. Originally, they were underpinned by the theory that trade flows determined foreign exchange rates. As the demand for a country s goods rose, so would the demand for the money to pay for them. Later, the idea that investment in a country sparked demand for its currency and determined exchange rates was used to support macroeconomic models. In both cases the models assumed that foreign exchange markets were efficient, information was public and market idiosyncrasies were irrelevant. They also assumed that traders preferences and expectations were homogeneous. Under these assumptions exchange rates would react instantaneously to new information, with the macroeconomic model s prediction of the new exchange rate becoming the new exchange rate. There would be no need for markets to discover the new rate by matching buy and sell bids. In fact, not even a single trade would have to take place at the new rate before it would be considered valid. Macroeconomic specifications could not be reconciled with their low predictive powers. They could explain only a fraction of the foreign exchange (FX) markets average volume of U.S.$1.5 trillion per day and could not account for the observed persistent and predictable excess return from FX speculation. Nor could macroeconomic models account for why a temporary deviation of the FX rate from the predicated rate one year would still affect the FX rate in the next year. 1 This paper draws heavily on Lyons (2001) and Sarno (2004). 2

3 The study of market microstructure models progressed in the 1990s as an attempt to resolve these puzzles. In his book The Microstructure Approach to Exchange Rates, Lyons (2001) tells how shortly after completing his doctorate he spent a day with a foreign exchange trader. Despite my belief that exchange rates depend on macroeconomics, only rarely was news of this type [the trader s] primary concern. Most of the time [he] was reading tea leaves that were, at least to me, not so clear. Those tea leaves were individual trades, the order in which they came and their bid-ask spreads. In market microstructure models these properties are analogous to detailed quantity and price data and help to reveal the individual participants beliefs about exchange rates. In this way market microstructure models fit with the theories of the Austrian economists who believed that the process of reaching equilibrium through pairing buy and sell offers uncovers important information about the participants knowledge. Because most trades come from professional traders who carefully analyze the market before placing their trades, this data is particularly valuable. Because these traders take macroeconomic variables into account when they analyze the market, microstructure models accommodate macroeconomic data, albeit through the filter of other traders. The evidence for market microstructure models is strong. They typically have an R 2 between 40 and 60 per cent, meaning they account for between 8 and 60 times more variance than macroeconomic models. They are also able to explain some of the puzzles that have confounded the macroeconomic models. For example, because market microstructure models reject the assumption that foreign exchange markets are efficient, they are able to resolve the paradox of excess returns. They assume that the FX markets are not transparent and 3

4 equate excess returns from speculation to traders at large financial institutions who see more trades and therefore possess more information than other traders. Market microstructure models explain the FX markets excess volume as FX traders using their own portfolios to fulfill customers orders and then passing these orders to other traders in order to rebalance their portfolios risk. The volume is generated as traders pass the orders from one to another like a hot potato. The long-run effect of a price deviation is explained by noting that any deviation contains information about the exchange rate and that this information has an ongoing impact on FX markets. Market microstructure models are not a repudiation of the idea that macroeconomic fundamentals drive exchange rates. Rather, they take a more nuanced view of macroeconomic data, giving it a higher weight once it has passed through the filter of traders. As microstructure models evolve they are finding a place for macroeconomic data alongside microstructure data and are leading us to a better understanding of FX markets. 4

5 Bibliography Lyons, Richard. The Microstructure Approach to Exchange Rates. Cambridge, Massachusetts: The MIT Press, Sarno, Lucio. Towards a Solution to the Puzzles in Exchange Rate Economics: Where Do We Stand. Working paper, University of Warwick,