ESRC Society Today

Standard macroeconomic models of exchange rate determination have proved themselves to be remarkably unsuccessful at explaining movements in foreign exchange (FX) rates in the short run.

As a result greater emphasis has been placed on microstructure factors such as ‘order flow’- net buying or selling pressure, to explain and help forecast exchange rates. Previously the lack of sufficiently large datasets has made it difficult to accurately asses the role of such variables on exchange rate determination. 

In this project, Professor Charles Goodhart and his research team from the London School of Economics, use large datasets from the Reuters D2000-2 FX trading system to test the microstructure hypotheses and extend understanding of exchange rate determination.

Key findings

Evidence on high frequency exchange rate determination

• Using a larger dataset and when examining a number of currencies, the previous order flow results are confirmed. Order flow does help determine exchange rates. 
• The informativeness of order flow is not constant but in fact varies depending on the prevailing market conditions, as described by spreads, return volatility and trading volume.      

Linkages between liquidity, transactions activity and volatility

• Using indicative quotes as proxies for firm/tradable quote prices is flawed. However, indicative quote frequency is a good proxy for actual/realised transaction frequency. 
• The Reuter’s D2000-2 trading system was found to be dynamically illiquid; after a market order, liquidity tends not to be restored on that side of the market. 
• After the introduction of the euro, spreads, as a percentage of mid-quote, increased but this was not caused by a fall in euro liquidity. Instead it was the granularity of pricing grids and the re-factoring of the exchange rate that caused percentage spreads to rise.     

Microstructure influences with lower frequency data

• Order flows can help explain exchange rates even when examining rates at lower frequencies.      

Interactions between microstructure variables and macroeconomic news

• Macroeconomic data that is released publicly and simultaneously to the market is incorporated into foreign exchange prices primarily through the trading process. 
• Traders tend to withdraw liquidity leading up to, and in the minutes immediately following, the releases of scheduled macro data, resulting in spreads widening and depth on Reuter’s D2000-2 system decreasing.     

Central Bank intervention and monetary policy

• Sterilised intervention does have significant and important effects on exchange rate returns. Such intervention is found to be more effective when performed in conjunction with other Central Bank intervention. 
• Sterilised intervention may on occasions be a better policy than tightening monetary policy in the face of speculative attacks on the currency. 
• After a crisis has hit, tightening monetary policy does not necessarily adversely affect other financial markets     

About the report  

The project used three large, tick-by-tick datasets of various exchange rates from the Reuters D2000-2 trading system and Olsen Data AG for the period from 1986 to 2000. The research has been disseminated through journals, economic conferences across Europe and at regular seminars at the London School of Economics.

Key words 

Exchange rate determination, Professor Charles Goodhart, London School of Economics

View all other award details