Carlos A. Ibarra
regressors. The presentation of results includes p-values for the J test
of adequacy of instruments and the Q test for serial autocorrelation.
Table 1 shows the main estimation results for Equation 3. All the
coefficients are statistically significant, with the exception of the
intercept and fed; these variables were kept, however, because without
them the model’s statistical performance was negatively affected. The
model passes the tests for absence of serial correlation (up to 36 lags)
and adequacy of instruments. Solving for the cointegration equation,
as in (4), yields:
(6) irdtc = -48.6077 + 34.0508 russia + 2.2697 tight - 3.6137 loose
+ 42.9631 lns - 13.4167 lnmt + 0.9093 πt + 0.3292 fedt.
Leaving aside the exchange rate, it can be seen that all the
coefficients have the expected sign. Thus, a tighter monetary policy
stance leads to wider interest rate differentials, and the same happens
with a rise in inflation. The size of the coefficients seems plausible.
For example, the long-run effect (i.e., after the dynamic effects have
been worked out) of a 10% rise in the real money base is a 1.3 point
fall in the interest differential; in the same way, a 1 point rise in the
inflation rate leads to a 0.9 rise in the interest rate gap.12
For our purposes, the most important result concerns the exchange
rate coefficient. Its estimated value implies that, holding everything
else constant, a 10% permanent depreciation eventually leads to a 4.3
point rise in the interest rate differential. Thus, we do not find the
negative relationship between the exchange rate and interest rates
frequently assumed in traditional models of monetary policy under a
float.13
As noted above, the short-run deviation from equilibrium is simply
υt = irdt - irdtc. The Phillips-Perron test statistic for this series (at the
Newey-West suggested truncation lag of 5) is -5.9123 with intercept
and -5.3042 without it, for sample size 319. The augmented Dickey-
Fuller test statistic (including 3 lags, as suggested by the Schwarz
12 The equation intercept implies that when the exchange rate, the real money supply, the
inflation rate and the Federal funds rate are at their mean values (2.1924, 3.5574, 14.3233 and
4.8634, respectively), the interest rate differential’s predicted value (setting all dummies at
zero) is 12.5; its actual sample mean value was 14.7.
13 It should be recalled that this result does not reflect exclusively the very short-run
interest rate-exchange rate nexus, but that it captures the dynamic response incorporating
almost six months of lagged impacts.
16