that the impact of financial liberalization on the UIP condition is ambiguous. In particular,
they report that the three Latin American countries in their study exhibit an increase in
the magnitude and volatility of excess currency returns after liberalization; whereas for the
five Asian countries and Turkey, excess returns are generally lower in the post-liberalization
period. Moreover, employing a three-factor Fama-French model (Fama and French, 1993)
with time-varying factors and betas, they conclude that the time-varying risk premium
differs significantly in the pre- and post-liberalization periods for all countries but one, with
no general pattern attributable to all countries. Splitting the sample in a pre-determined
way, Mansori (2003) explores whether the introduction of euro and the adoption of accession
partnerships with the EU have an effect on the UIP condition for the Central European
economies. His findings suggest that the UIP condition holds for the period 1994-2002, and
the analyzed structural breaks do not seem to matter.14
Previous empirical studies take the timing of the regime-switch in a pre-determined way
by using the official declaration date as given. Although setting a regime switch date for
estimation purposes is unescapable, as Bekaert et al. (2002) point out, de facto date of
structural break does not need to coincide with the official declaration date. In particular,
Bekaert et al. (2002) study the timing of regime switches in 20 emerging market economies
for the period 1980-1996 with a focus on equity markets and show that the endogenous
regime switch date occurs mostly within 3 years of one of the major liberalization dates.
Goh et al. (2006) recognize the importance of determining the regime switch date
endogenously, and incorporate this in a UIP framework. They study Malaysia for the period
1978-2002, which covers episodes of interest rate liberalization, currency crisis, exchange rate
controls and economic recessions, making Malaysia a good candidate for such an analysis.
Particularly, they employ the switching ARCH model of Hamilton and Susmel (1994) in
which the regime change is analyzed through endogenizing the timing of the regime change.
The regime change is taken as shifts in the conditional volatility of deviations from the UIP
condition over different regimes and the timing of the regime switch is estimated through
a first-order Markov process. Goh et al. show that deviations from the UIP condition are
13