Efthymios G. Pavlidis, Ivan Paya and David A. Peel 1027Average daily global turnover in foreign exchange market transactions was esti-
mated at around $2.7 trillion in 2006. In an efficient speculative market, prices
should fully reflect publicly available information so that it should not be possi-
ble for a trader to earn systematic abnormal returns. However, empirical estimates
show that the spot exchange rate next period moves on average in the opposite
direction to that currently predicted by the forward premium, the so-called forward
bias puzzle. In section 22.3.1 we consider econometric tests of the covered interest
parity condition and, in section 22.3.2, the uncovered interest parity condition
and the various explanations of the forward bias puzzle.
If expectations are forward looking then the exchange rate regime in place, as
well as the anticipation of the implementation of future exchange rate regimes, will
impact on the behavior of the exchange rate. An important example of such policy
arrangements are “target zones.” Within this framework the authorities intervene
to attempt to keep the exchange rate within a band. The recent experience of
European currencies between 1979 and 1998 under the European Monetary System
(EMS) is one such example.^1 In section 22.4 we consider some of the econometric
testing of target zone models.
As with other asset markets, researchers have examined whether exchange rates
exhibit rational speculative bubbles.^2 In section 22.5 we briefly discuss some of the
more recent developments in the area.
Section 22.6 covers the issue of exchange rate forecasting. In a seminal paper,
Meese and Rogoff (1983a) compared the out-of-sample forecasts produced by vari-
ous exchange rate models with forecasts produced by a random walk (RW) model.
On the basis of the root mean square criterion, they concluded that none of the
various asset-market exchange rate models they considered outperformed a simple
RW. Since then a plethora of papers have been published on this issue. We give an
overview of developments. Finally, section 22.7 provides a brief conclusion of the
major issues to emerge in this chapter.
22.2 Real exchange rates
The PPP hypothesis states that domestic prices in two countries should be the same
when converted to a common currency. Let us define the log real exchange rate as
yt=st−pt+p∗t, wherestis the logarithm of the spot exchange rate (the domestic
price of foreign currency),ptis the logarithm of the domestic price level andp∗t
the logarithm of the foreign price level.
Initial empirical studies of PPP consisted of fitting a univariate autoregressive
model for the real exchange rate,yt=
∑p
i= 1 βiyt−i+εt. In the 1980s many empirical
studies were unable to reject the unit root null hypothesis forytusing standard
unit root tests such as the augmented Dickey–Fuller (ADF) and Phillips–Perron
(PP) (Taylor, 1988; Mark, 1990). In the light of these results, the first econometric
issue addressed in the literature was the power of the ADF and PP tests in a linear
framework. Lothian and Taylor (1996) undertook the following simulation. They
generate a stationary AR(1) process calibrated with empirical estimates on data for
100 years of the dollar–pound real exchange rate asyt=0.87yt− 1 +εt.^3