The US Dollar/Euro Exchange Rate: Structural Modeling and Forecasting During the Recent Financial Crises

• Published date: 01 December 2017
• Author: Claudio Morana
• Date: 01 December 2017
• DOI: http://doi.org/10.1002/for.2430

Journal of Forecasting,J. Forecast. 36, 919–935 (2017)

Published online 21 July 2016 in Wiley Online Library (wileyonlinelibrary.com)DOI: 10.1002/for.2430

The US Dollar/Euro Exchange Rate: Structural Modeling and

Forecasting During the Recent Financial Crises

CLAUDIO MORANA1,2

1

Dipartimento di Economia, Metodi Quantitativi e Strategie di Impresa, Università di

Milano–Bicocca, Milan, Italy

2

Rimini Centre for Economic Analysis, Italy

ABSTRACT

The paper investigates the determinants of the US dollar/euro within the framework of the asset pricing theory of

exchange rate determination, which posits that current exchange rate fluctuations are determined by the entire path of

current and future revisions in expectations about fundamentals. In this perspective, we innovate by conditioning on

Fama–French and Carhart risk factors, which directly measures changing market expectations about the economic out-

look, on new financial condition indexes and macroeconomic variables. The macro-finance augmented econometric

model has a remarkable in-sample and out-of-sample predictive ability, largely outperforming a standard autoregres-

sive specification. Wealso document a stable relationship between the US dollar/euro Carhart momentum conditional

correlation (CCW) and the euro area business cycle. CCW signals a progressive weakening in economic conditions

since June 2014, consistent with the scattered recovery from the sovereign debt crisis and the new Greek solvency

crisis exploded in late spring/early summer 2015. Copyright © 2016 John Wiley & Sons, Ltd..

KEY WORDS US dollar/euro exchange rate; asset pricing theory of exchange rate determination; macroeco-

nomic and financial determinants; risk factors; subprime mortgage financial crisis; sovereign debt crisis;

early warning indicators of macroeconomic and financial stress; forecasting; multivariate GARCH

model

INTRODUCTION

Since its introduction in January 1999, the euro has rapidly gained the role of numeraire, medium of exchange

and store of value for international transactions, even challenging the hegemony of the US dollar as predominant

international currency at a certain extent. The bilateral US dollar/euro exchange rate is currently the most relevant

currency pair in the foreign exchange market. Accurately forecasting the US dollar/euro is then important for both

practitioners and policymakers. Wieland and Wolters (2013), for instance, show that Central Bank policies in the USA

and Europe are well described by interest rate rules, where interest rates are set according to forecasts of inflation and

economic activity. By influencing current account projections, as well as inflation and gross domestic product (GDP)

growth predictions eventually, forecasting the US dollar/euro exchange rate is then of utmost relevance for the proper

conduct of economic policy.

Surprisingly, little is known about the structural determinants of US dollar/euro exchange rate, particularly during

the subprime mortgage and sovereign debt financial crises, as well as the post-crisis period. Most of the evidence is in

fact available for the pre-crisis period. For instance, Sartore et al. (2002) consider a structural econometric model for

the real US dollar/euro exchange rate in VECM form, using data from 1990 to 1999. They show that real long-term

interest rate spreads, foreign trade efficiency measures, fiscal policy differentials and commodity prices are relevant

determinants of the currency. Other studies have provided evidence of convergence towards purchasing power parity

(PPP) in the euro area. In particular, Lopez and Papell (2007), using panel data methods, find evidence of PPP within

the eurozone and between the eurozone and its main partners over the period 1972–2001; similarly, Koedijk et al.

(2004) and Schnatz (2007) over the periods 1973–2003 and 1980–2006, respectively.1

The validity of the monetary model of exchange rate determination has also been directly assessed in recent studies.

For instance, Nautz and Ruth (2008) find a theoretically coherent response of the nominal US dollar/euro exchange

rate to EA and US monetary disequilibria over the period 1981–2005. Chen et al. (2011) investigate cointegration of

the US dollar/euro with fundamentals posited by the monetary exchange rate model, such as money stocks, prices and

real output, over the period 1994–2003. They show that both short-run (price stickiness) and long-run (secular growth)

Correspondence to: Claudio Morana, Dipartimento di Economia, Metodi Quantitativi e Strategie di Impresa, Università di Milano–

Bicocca,Milan,Italy. E-mail: claudio.morana@unimib.it

1Evidence of nonlinearity is also pointed out by Schnatz (2007), as the speed of mean reversion in the US dollar/euro exchange rate would rise

nonlinearly with the magnitude of the PPP deviation. See Camarero and Ordonez (2012) for further evidence of nonlinearity in mean reversion,

yet toward the fundamentals represented by the productivity differential, overthe period 1970–2009.

Copyright © 2016 John Wiley & Sons, Ltd.

920 C. Morana

fundamentals affect the currency.Other studies have documented the superior forecasting accuracy of hybrid versions

of the monetary model. For instance, Chinn and Moore (2009) investigate the period 1999–2007 and augment the

monetary model fundamentals with order flow variables, consistent with the Evans–Lyons microstructure approach,

in order to account for the potential role of innovations in public and private information. Beckmann et al. (2011)

focus on the period 1975–2007 and augment the monetary model with imbalance measures, such as the tradables over

non-tradables price ratio and the trade balance, consistent with Harrod–Balassa–Samuleson effects and the portfolio

model of exchange rate determination. Superior forecasting accuracy is also shown by Dal Bianco et al. (2012) for

weekly US dollar/euro returns over the period 1998–2010, using a mixed-frequency econometric model based on

monetary fundamentals quoted at the weekly and monthly frequencies.

The lack of empirical evidence for the recent subprime and sovereign debt financial crises surely is an important

gap in the literature, which the paper then aims to fill. In particular, we estimate a reduced-form econometric model

for both the conditional mean and variance of the US dollar/euro exchange rate return, conditioning on a large set of

macroeconomic and financial variables. The econometric model is grounded on the asset pricing theory of exchange

rate determination, which posits that the current exchange rate depends on the present discounted value of the future

stream of fundamentals. In the latter framework, exchange rate fluctuations are then accounted by news or revisions

in expectations about fundamentals. We then innovate the literature by including in the information set, in addition to

standard macroeconomic fundamentals, financial condition indexes and direct measures of changing market expecta-

tions about the economic outlook, as yield by risk factors, such as the five Fama and French (1993, 2015) factors and

Carhart (1997) momentum, consistent with their usual interpretation in terms of proxy for state variables, capturing

changes in the investment opportunity set. A strict linkage between risk factor shocks and economic dynamics has

also been documented by Morana (2014b), consistent with the recent strand of ‘news-driven’ business cycleliterature,

which has drawn attention to the role of abrupt changes in expectations in driving economic fluctuations (Beaudry

and Portier, 2014).

To preview some of the results of the paper, we find that the estimated hybrid monetary model (AR-MF) shows a

remarkable in-sample predictive ability, accounting for 60–80% of US dollar/euro returns variance—fivefold larger

than for standard autoregressive models (AR) neglecting macro-financial information. This is not due to overfitting,

as the AR-MF model yields an average 30% reduction in root mean squared forecast error (RMSFE) out-of-sample

and is forecast encompassing relative to the AR model.

Moreover, by assessing linkages in second-moments, we uncovera stable relationship between the US dollar/euro–

Carhart momentum conditional correlation and the state of the EA economic cycle. A progressive weakening in EA

economic conditions is then detected since June 2014, consistent with the sluggish and scattered recovery from the

sovereign debt crisis which eventually led the European Central Bank (ECB) to introduce the quantitativeeasing (QE)

policy in January 2015.

The rest of the paper is organized as follows. In the next section we provide theoretical insights on the role of risk

factor information in exchange rate determination. In the third section we estimate a reduced-form econometric model

for the US dollar/euro exchange rate and assess its forecasting performance over three different time spans, covering

the subprime mortgage crisis, the sovereign debt crisis and the post-sovereign debt crisis period, respectively. In

the fourth section we assess second moment linkages between US dollar/euro exchange rate and risk factor returns.

The fifth section concludes. Additional details about data, methodological contributions and empirical results are

contained in the online Appendix (supporting information).

EXCHANGE RATE DETERMINATION AND RISK FACTORS

Consider the asset pricing model for exchange rate determination:

etDQ

Mt˛ŒEtetC1et(1)

where etis the (log) exchange rate eUS$=Cat time period t, i.e. the value of one unit of local currency (C) in foreign

currency units (US $), EtetC1is the exchange rate expected at time tC1based on time tinformation, ˛>0is a

coefficient and Q

Mtrepresents the fundamentals at time period t.

Rearranging, one has

etD1

1˛Q

Mt˛

1˛EtetC1(2)

Forward iteration of equation (2), under the assumption of rational expectations, then yields

etD1

1˛

1

X

D0˛

1˛

EtQ

MtC(3)

showing that the current exchange rate depends on the present discounted value of the future stream of fundamentals,

where the discount rate is ˛.

Copyright © 2016 John Wiley & Sons, Ltd. J. Forecast. 36, 919–935 (2017)