Order flow analysis of exchange rates.

Author:Rime, Dagfinn

Norges Bank recently started to collect new foreign exchange statistics. (2) These statistics provide ah overview of which foreign currencies various market participants buy and sell against NOK. Participants' purchases and sales (order flow) are an important variable in exchange rate analysis using microstructure models. Order flow analysis has proved to be helpful in understanding changes in the exchange rate in the short term and the new statistics are well suited for this type of analysis. This article provides an overview of the theory behind order flow analysis and uses the data collected to date to illustrate some concepts.

1 Why order flow analysis?

According to economic theory, exchange rates ought to be determined by a number of macroeconomic conditions. The theory of purchasing power parity posits that the level of the exchange rate between two countries should be the same as the relative price level between the two countries. The theory of uncovered interest parity postulates that the exchange rate today should not systematically deviate from the differential between the exchange rate and interest rate 'some time ago' (depending on the maturity of the interest rates). When combined to create a macro model, for example the Mundell-Fleming model, exchange rates are also determined by GDP growth. Interest rates, inflation, and economic growth are often called macro fundamentals.

Empirical studies show that macro fundamentals can explain movements in the exchange rate relatively well, particularly over longer time horizons such as six months or a year, but their explanatory power is lower for daily or weekly horizons. Sometimes, exchange rates seem to live a life of their own, as if completely detached from macro fundamentals. (3)

Exchange rate deviations from fundamentals can be substantial and persist over a sufficiently long period to be significant. What causes such deviations and why do macro fundamentals not 'function' in the short term? This article discusses exchange rate determination in the short term: hence, why exchange rates may deviate from what is believed to be the macro equilibrium exchange rate. Order flow analysis has proved to be useful in establishing this connection and the discussion is therefore based on the theory underlying order flow analysis: the theory of financial market microstructure. Microstructure theory looks at participants in the market and the constraints they face. The application of microstructure theory to the foreign exchange market is a relatively new field of research (late 1990s), and the main contribution to date has been provided by focusing on possible differences in participants" expectations.

That expectations regarding securities prices differ is not new. Insiders in the stock market have been studied for years. Different expectations in the foreign exchange market may, however, seem slightly odd. After all, vast empirical research and the bulk of theory show that exchange rates are determined by macro fundamentals in the long term. Is it not the case that macro fundamentals can be equally well observed by all market participants all the time, and thereby pin down homogeneous expectations? We will therefore look more closely at what might give rise to different expectations in the exchange rate market and how this is captured by order flows. The theory will also be illustrated by an empirical analysis based on the data reported to Norges Bank in the new foreign exchange transaction statistics (see Meyer and Skjelvik, 2006).

2 Different expectations = different information?

When discussing possible sources of differences in expectations, the following expression of the exchange rate may be useful. An exchange rate is determined by:

[P.sub.t] = E[[P.sub.l+1] ([F.sub.t+1])|[[??].sub.t] / l+[r.sub.t]+[[rho].sub.t] (1)

where P is the exchange rate (e.g. NOK per EUR), which is a function of expectations regarding discounted future macro fundamentals F and the information set [??], at time on which these expectations are based. (4) E is the expectations operator, r is the interest rate and [rho] is a risk premium. The equation implies that today's price is the discounted value of tomorrow's expected price, where tomorrow's expected price is determined by the information available today and anticipated changes in macro fundamentals.

Determining what the correct value of the exchange rate is today is ah extremely difficult task because there is so much information that market participants ought to know: What is today's GDP? Or what is the rate of inflation? Who knows what is 'expected" or what that might imply for GDP or inflation tomorrow, or in a month's time? Difficult? Let us continue: Who knows what the correct discount rate is for exchange rates? And to make things even more difficult: how can you know how exchange rates will react to macro fundamentals (the functional form in the expectation, the 'correct' model), when there is insufficient supporting empirical evidence? Finally: Who knows which information set can be used to answer these questions?

A number of disappointing empirical results show that few people are fortunate enough to have the answers to all these questions (see overview articles mentioned in footnote 3). And yet every day, market participants have to base their decision on some form of information when setting their prices (market-makers) or taking a position (investors). We will look more closely at how they might make this decision in the next section and focus on possible reasons for differing expectations in the remainder of the section.

First we would like to clarify one question. Given that it is so difficult to determine the exchange rate, are different expectations necessarily the best way to explain why movements in the exchange rate deviate from what is indicated by macro fundamentals? Understandably, it is difficult to determine the fundamental exchange rate when everyone has equal access to insufficient and uncertain information. But, if participants have rational expectations, they should not make systematic errors, which they seem to do in the short term in the foreign exchange market (when not using the information set used by the market itself). It might be that not all market participants have rational expectations. If that is the case, it seems reasonable to assume that they would also have different expectations. Microstructure theory is based on rational...

To continue reading