|
In this paper, the linearity and stationarity of the real
exchange rates of India, Nepal, Pakistan and Sri Lanka are
investigated using formal linearity and the recently developed
nonlinear stationary test procedures. Results obtained show
that these real exchange rates are stationary despite the
presence of nonlinearity. This finding suggests that, it
is possible to monitor and forecast the behavior of these
nominal exchange rates, as well as to determine their equilibrium
values using the corresponding relative prices.
Numerous documentations on the findings of nonlinearity
in the exchange rates have been recently added to the existing
exchange rate study. Among others, Micheal et al. (1997),
Sarantis (1999), Taylor and Peel (2000), Baum et al. (2001)
and Peel et al. (2001), have reported the existence of nonlinear
exchange rate behavior in the context of developed nations.
Earlier, Peel and Speight (1996) had detected nonlinearities
in the exchange rate of East European countries. In a separate
endeavor, Ma and Kanas (2000) found nonlinearities for countries
under Exchange Rate Mechanism. Sarno (2000a and 2000b),
on the other hand, documented the presence of nonlinearity
in the real exchange rates of Middle East and highly inflation
countries. Of late, Liew et al. (2003 and 2004) and Liew
(2004) found strong evidence of nonlinear behavior of US
dollar as well as Japanese yen-based real exchange rates
in the Asian region. This is followed by Anoruo et al. (2006),
that complements the literature by offering empirical evidence
of nonlinear real exchange rates from the African continent.
One important implication of these documentations is that
linear testing frameworks may no longer be taken for granted
as adequate tools in the study of exchange rate. Another
equally crucial implication is that linearity property of
exchange rates, which has been neglected in the past, partially
due to ignorance of the plausible presence of nonlinearities
and partially due to the unavailability of advance information
and computer technology, must be predetermined using formal
linearity test prior to the application of any econometric
testing and estimation procedures. Otherwise, robustness
of the results and relevance of the inference made from
these studies are doubtful. Conventionally, it has been
a common and formal practice to subject time series including
exchange rates to linear testing and estimation procedures
with the unjustified assumption that the series being tested
is linear in nature. Remarkably, these results are meaningful
only if the null hypothesis of linearity has not been rejected
by formal linearity test (Liew et al., 2003; and Tang et
al., Forthcoming). In this respect, one easily conducted
the formal linearity test, the Luukkonen, Saikkonen and
Teräsvirta (LST) linearity test (Luukkonen et al.,
1988), which has been adopted in most of the above-mentioned
studies to uncover the evidence of nonlinearity in the real
exchange rates. Remarkably, besides exchange rate study,
the usefulness of LST test has been extended to the study
of income convergence (Liew and Lim, 2005) and balancing
item (Tang et al., Forthcoming).
|
