In order to access the short run relationship between a shock in one specific variable and
the responses of the others, it was conducted an analysis of impulse response functions.
Given the cointegration relationship identified in the previous section, a vector error
correction (VEC) model was estimated. Again, using the multivariate AIC in a general-to-
specific search method, the order 1 was found as the best specification for the VEC system.
The VEC(1) was estimated using the first differences of the variables and including the first
lag of the error correction mechanism for the cointegrating vector of the interest rate
equation. The restrictions imposed on the VEC to identify the structural residuals followed
the Choleski decomposition, with the order of the variables assumed to be gapt, cpit, it, and
rdpt. The motivation is that gapt is considered the most exogenous and rdpt the most
endogenous variable in the model.
5