In order to test for robustness of our results, we perform alternative
estimations considering (i) different computations of the business cycle
(see Output gap in Tables 2a and 2b), (ii) different measures of the long-run
level for the monetary policy rate (Table 2a, last two columns, and Table 2b,
last three columns), and (iii) different instruments, like the terms of trade
and the U.S. output gap (Tables 2a and 2b, Column 6). Results are virtually
invariant (the range for the threshold level is estimated a t 222-286 bp15)
and, consequently, the main conclusions are supported.