The sign of the error correction term, EC(-1), in the final specifications is crucial. This
coefficient reflects the self-correcting dynamic mechanism of the error correction model, and is
expected to be negative. If the fundamentals in the last period dictate a lower real exchange rate
than that observed, then the real exchange rate will strictly depreciate in the current period. The
sign is negative in all the specifications. Stability requires the absolute value of the coefficient to
be between zero and one, and this is the case in all specifications.
In all cases the fundamentals that were found to be significant in the long run equilibrium
relations (Tables 5 and 6) are also found to be significant in the error correction equations
(Tables 7 and 8), except TOT for Russia. In general, the original signs are retained, except for
FLOW, OPEN, and GOV for Poland. Thus, the ‘temporary’ effects of the fundamentals on the
short run movements of the real exchange rates are important -as Elbadawi (1994) suggests.
NEER and DCRE in the error correction equations account for the effects of nominal devaluation
and expansionary macroeconomic policy. Nominal devaluation is statistically significant with a
negative impact on the short run movement of the real exchange rate, as expected. The
expansionary macroeconomic policy, which is expected to appreciate the real exchange rate, is
also found to be significant for both countries.
EGARCH in the final specifications has special significance. The error terms are not only
serially correlated, but also have non-constant volatilities that are conditional on the past squared
error terms. This means large swings of the real exchange rate in the current period can be at
least partly explained by its past volatility. This may happen in spite of the presence of the error
correction term in the short run error correction equation for two reasons. First, the strength of
the error correction term may not be large enough to fully correct the exchange rate quickly (note
that in the estimated error correction equations none of the related coefficients is close to one in