The passthrough of exchange rate changes to domestic inflation is a key issue in the
Phillips curve set-up. Several linear and non-linear specifications for the passthrough coefficients
have been tested, reducing to four the alternatives implemented in the preferred simulation tool.
The first one is a standard constant coefficient; simply estimated from a suitable sample of past
data. The second one is a quadratic transfer from exchange rate variations to inflation. The third
one is a level-dependent coefficient. It is estimated under the assumption that the passthrough
depends also on the level of the (log) nominal exchange rate. The last one is a quadratic function
of the nominal exchange rate level, motivated by a simple partial equilibrium model in which
exchange-rate devaluations shift the supply curve of competitive producers of tradable goods.10
All non-linear variants intend to capture more precisely the effects of a temporary exchange rate
overshooting.11 For the small number of observations available in a quarterly frequency, however,
their results were very close to the linear variant and consistent with international evidence that
the passthrough coefficient is inversely proportional to the degree of real exchange rate
appreciation at the moment prior to the devaluation. The equations below summarize the four
alternative specifications.