Liquidity and Uncertainty Dependent Order Flow Effects in the Foreign Exchange Market: an Empirical Investigation : A Microstructure Approach to Exchange Rate Modeling

Abstract

In this thesis I try to model exchange rate fluctuations using microstructure theory. While the classical macroeconomic models of exchange rates have shown to have little or no support at frequencies higher than semiannual (Rime and Sojli, 2006), microstructure models have proved successful in explaining exchange rate movements in relative high frequent data (daily and weekly frequency). Recent literature have attempted to apply microstructure in order to explain exchange rate movements at longer horizons in order to fill the hole which macroeconomic models have proved unable to explain. Chinn and Moore (2010) used monthly data to show that order flow does have explanatory power for exchange rate fluctuations at this level as well. The authors also found cointegrating relationship between order flows and exchange rates, confirming that the effect of order flow is persistent.

This thesis is an extension to the work done at this level of frequency. I utilize two microstructure models in order to explain monthly exchange rate fluctuations for four major exchange rates. Order flows are created from foreign exchange trade statistics published in the Treasury Bulletin. Since literature and theory have shown that the price impact of order flow is dependent on market conditions, I try to control for this variability by adjusting the order flows. Building on theoretical foundations explained in Killeen et al. (2006), I therefore choose to adjust the order flows for uncertainty and liquidity. Furthermore, an attempt to control for hedging by including net option positions into the models is also provided. The Johansen trace test is applied in order to test for cointegration relationships between order flows and exchange rates.

The results obtained in this thesis show that adjusted order flow provides better in-sample fit compared with unadjusted order flow. The results also points towards an “undershooting” adjustment mechanism for the exchange rates as the long run price impact are larger than the short run impact.

The thesis is organized in following manner; chapter 2 outlines two different approaches to exchange rate modeling, the classical way with macroeconomic “fundamentals” and the microstructure approach which highlights agents heterogeneity and information superiority. The role of order flow in this setting is explained and also why it is an important determinant for explaining exchange rate fluctuations. Also included are some important facts about the exchange rate market structure. Chapter 3 contains a brief, but intuitive explanation of the two models used in this thesis; the Portfolio Shift model and the Liquidity Shock model. Chapter 4 gives a thorough description of the data series used in this thesis. An explanation of how I create order flow series is also provided. Furthermore, the times series properties for the dataset are tested for and discussed. Chapter 5 presents the results found in this thesis. Section 5.1 deals with econometric results from the Portfolio Shift model with various sorts of proxies for order flow. Section 5.2 show results from Johansen cointegration testing for both the models. Vector Error Correction Models based on the Liquidity Shock model are also estimated for each of the exchange rates. Chapter 6 concludes and proposes some potential extensions to this thesis.

Estimations and other econometric results in this thesis have been done with the statistical software EViews, version 7.1.