Carlos A. Ibarra
structure in the regression analysis (an aspect that Equation 1 does
not explicitly capture). It could very well be that in the very short run
a rise in the exchange rate leads to higher interest rates, but that over
time this bandwagon effect loses strength, and eventually a fall in
expected depreciation lowers interest rates.
b) Specification
As could be expected, for most of the series under analysis it is not
possible to reject the hypothesis that they contain a unit root. Thus,
the econometric analysis will proceed in two steps, the first being the
estimation of a cointegration equation from an autoregressive
distributive lag (ADL) model for the level of the interest rate
differential; the second step involves the estimation of an ECM in
differences, from which an impulse response function for the interest
differential can be derived.
Our purpose is to estimate the dynamic response of private
expectations and risk assessments,8 as reflected in the evolution of
the peso-dollar interest rate differential, following a permanent rise
in the exchange rate. This differential may of course also reflect
monetary policy actions that affect asset supplies and local liquidity
conditions. Thus, the regression equation includes, as a control variable,
an index of the real money supply divided by a measure of economic
activity; it also includes the inflation rate as the main determinant of
the trend interest differential, and the Federal funds rate as an
indicator of the US monetary policy stance.
The starting point for the estimation of the cointegration equation
was the following ADL model for the interest rate differential:
(2) irdt = a0 + ς aj irdt-j + ς bi lnst-i + ς ci lnmt-i + ς fi πt-i + ς gi fedt-i + vt,
where j = 1, 2, ..., L, i = 0, 1, ..., L, L is the number of lags included in
the equation, t denotes the week, irdt-j the interest rate differential,
lnst-i the log exchange rate, lnmt-i the log adjusted real money base,
πt-i the annual inflation rate, fedt-i the Fed funds rate, and vt an error
8 Berg and Borensztein (2000) discuss why a country’s political risk premium may be
positively correlated with expected depreciation, and illustrate with data from Argentina. See
also Edwards (2001), Figure 4.
12