The Interest Rate-Exchange Rate Link in the Mexican Float
nominal base divided by the product of the consumer price index and
the monthly industrial production index calculated by the National
Institute of Statistics (INEGI).
The sample runs from the last week of June 1996 to the last week
of July 2002. The initial specification had a long lag structure (24
lags), with the purpose of capturing the protracted effects of exchange
rate variations on the interest differential. This initial structure was
simplified according to variations in the Schwarz and Akaike criteria.
In the end, 22 lags were included for each variable.
During the period under analysis, monetary policy in Mexico was
conducted under a system of zero average (over 28-day periods) reserve
requirements for commercial banks (see Yacaman 1999 for a detailed
description). The Banco de Mexico is committed to satisfy whatever
level of reserve demand comes from commercial banks. But the
conditions under which such reserves are created are indicative of the
central bank’s policy stance. In particular, if the Banco de Mexico
announces a so-called “corto”, or shortage of some amount, it means
that such volume of commercial bank reserves will be supplied at
penalty rates. A rise in the “corto” has the purpose of pushing up market
interest rates since it induces banks to compete for funds so as to
avoid the penalization. In fact, it has been documented that such action
does have a very short-run, transitory impact on the Cete interest
rate (see D^az de Leon and Greenham, 2000).
Our regression equations include two variables intended to capture
this effect; in particular, tight is a dummy that equals one in the three
weeks following a rise in the short, and zero in the rest, while loose is
equal to one if the observation falls within three weeks after a reduction
in the short, and zero otherwise.11 There is also a dummy that captures
the (immediate) impact of the fall of 1998 financial market turbulence
linked to the Russian debt default; thus, russia equals one during the
first three weeks of September 1998, and zero in the remaining of the
sample.
It can be reasonably assumed that the interest rate differential,
the exchange rate, and the real money supply are affected by common
shocks. Thus, both the cointegration equation and the ECM were
estimated by GMM to allow for the possible endogeneity of the
11 The rationale for introducing these dummies is as in the inclusion of real money: to
isolate the effect of exchange rate variations on interest rate differentials, keeping constant
conditions in the money market.
15