FX Trading and the Exchange Rate Disconnect Puzzle

Abstract

This paper examines how trading in the FX market carries the information that drives movements in currency prices over minutes, days and weeks; and now those movements are connected to interest rates. The paper first presents a model of FX trading in a Limit Order Book (LOB) that identifies how information from outside the market is reflected in FX prices and trading patterns. I then empirically examine this transmission process with the aid of a structural VAR estimated on 13 years of LOB trading data for the EURUSD, the world's most heavily traded currency pair. The VAR estimates reveal several new findings: first, they show that shocks from outside the LOB affect FX prices through both liquidity and information channel; and that the importance of these channels varies according to the source of the shock. Liquidity effects on FX prices are temporary, lasting between two and ten minutes, while information effects of shocks on prices are permanent. Second, the contemporaneous correlation between price changes and order flows varies across the shocks. Some shocks produce a positive correlation (as in standard trading models), while others produce a negative correlation. Third, the model estimates imply that intraday variations in FX prices are overwhelmingly driven by one type of shock, it accounts for 87% of hour-by-hour changes in the FX prices. The second part of the paper examines the connection between the shocks in the trading model and the macroeconomy. For this purpose, I use the VAR estimates to decompose intraday FX price changes and order flows into separate components driven by different shocks. I then aggregate these components into daily and weekly series. I find that one component of daily order flow is strongly correlated with changes in the long-term interest differentials between US and EUR rates. This suggests that the intraday shocks driving this order flow component carry news about future short-term interest rates which is embedded into FX prices. I find that intraday shocks carrying interest-rate information account for on average 56% of the variance in daily EURUSD depreciation rate between 2003 and 2015, but their variance contributions before 2007 and after 2011 are over 80%. These findings indicate that the EURUSD depreciation rate is relatively well-connected to macro fundamentals via a particular component of order flow. Finally, I show that flows embedding liquidity risk have forecasting power for daily and weekly EURUSD depreciation rates.