Inflation Level. Table 8 presents the estimated ATTs of adopting an inflation targeting policy on the level of inflation. The first two columns specify the parameters (β and α) under which the hy- perbolic mean is calculated (here β > 0), and the remaining columns respectively show the estimated overall effect, the effect after 5 and 10 years of targeting. As seen on the table, we obtain ATTs whose negative sign is robust to the matching tools and variations of the hyperbolic mean parameters. This is an implication that on average the level of inflation for Targeters has been lower at the mean than that of Non-Targeters over the chosen sample period. However, important ATT differences still exist between our score functions and PS. Propensity Score matching, for instance, yields estimates that are both smaller and mostly statistically insignificant (first row of the Table 8). Such estimates could prompt one to argue for a very subdued and largely insignificant contribution of inflation targeting in the chosen sample. However, this clearly contrasts with results obtained via our various hyperbolic means, where 49 out of 54 estimates are significant at the 1% level, 4 at the 5% level, 1 at the 10% level, and only 1 is statistically insignificant, indicating a strong influence of IT in lowering inflation levels for Targeters. One interpretation is as follows: The linear index function specifies infinite substitution between components of X! They are effectively “the same” as far as characterizing the economy. When this level of substitution is made finite, with other values of β, different influences are exerted by different components of the economic variables in X, aside from the weights attached to each. This seems to raise the ATT for a wide range of the substitution and weight parameters. It also suggests that the corresponding implicit values of these same parameters may be close to the boundary sets, since they do not contradict the general inference of an average reduction in inflation by targeting.
