Information in foreign exchange markets

Abstract

This thesis investigates the theory that customer order flow conveys information about foreign exchange (FX) prices. This proposition has not been extensively tested due to the difficulties in obtaining sufficiently detailed private trading data. I use a unique data set from a leading Australian commercial bank that has records on every FX trade made in the spot AUD/USD market by the bank between 2005 and 2011. To address the research question, I first search for cointegration between the FX rate and cumulative order flow. This is a testable implication of the Evans and Lyons (2002b) model. I find little evidence in support of a cointegrating relation. Instead I turn to a static analysis on order flows and returns using simple regression. The results find no statistically significant correlation between customer order flow and FX returns. However, consistent with the liquidity provision role of non-financial clients in Evans and Lyons (2002b), I find a statistically significant negative correlation between order flow from the Diversified economic sector and FX returns. Next, I perform a dynamic analysis on order flows and returns using a Vector Autoregression model. The results suggest that order flow has a limited price impact with the exception of the Diversified sector. This sector's price impact tends to be negative, which is consistent with the static analysis. The variance decompositions show evidence of significant information shares, which is consistent with previous findings showing that order flow can be informative about the FX rate without having a permanent price impact. Finally, I focus on individual clients and use Monte Carlo simulation to generate a benchmark for expected cash flows that allows me to isolate potentially informed clients based on trading profitability. The results find few exceptionally profitable clients, consistent with the weak information effects found in the other chapters. These results suggest that the customer order flow of a typical commercial bank, without a significant institutional market share, does not carry information about FX prices. Even if customer order flow is informative about the exchange rate, its practical application requires data on a large proportion of market-wide institutional trading. The benefit of FX microstructure to smaller market participants appears limited. Show more