Wednesday, March 6, 2013

Monetary stimulus in high-inflation regimes

In countries where inflation is high and highly variable, you would not expect that monetary policy is optimal in a social-welfare sense. After all, such inflation is a sign of political influence, and from politicians that do not have the good of the people in mind: strongly expansionary monetary policy before elections, financing government expenses with a hidden inflation tax instead of other visible taxes, and even just pocketing monetary injections. Still suppose that for some exogenous reason inflation is high and highly variable, is their still scope for optimal monetary policy (beyond working toward a long-term goal of getting this under control)?

Wojciech Charemza, Svetlana Makarova and Imran Shah claim that yes, the central bank still can do good. They claim that output can be stimulated with a monetary stimulus when inflation expectations are much higher than output-neutral inflation. The latter is obtained with a two variable VAR, and then the sign of the residuals is used to interpret asymmetric impulse responses. Like many VAR analyses, this is a little bit voodoo science, especially when you look at countries whose data records are poor if not manipulated. Of course, following this policy will not help in solving the inflation problem of these economies (defined as at least 4.8% inflation at least 25% of the time). But if expectations are much lower, then one can tighten monetary policy without big consequences. Unfortunately, a VAR cannot tell us why all this would be happening, only that there is a statistical coincidence.

1 comment:

VARman said...

Var..., voodoo science...? Perhaps yes, but where is the alternative? Take the DSGE models for example. If VAR modelling is voodoo, DSGE is ivory tower: plenty of shiny, elegant assumptions in which you have to believe strongly notwithstanding the data. If data do not fit to the theory, never mind, the model must be good, even if not testable. VAR's, for that measure, usually fir to the data and explain the past dynamics.

By the way, it seems that authors look at the differences between the expected and output neutral inflation (and the trick is in here, as most people think these are the same) and build three-equation VAR's on the output and two inflation series, split by the sign of this difference. They did not quite look at the signs for the residuals while looking for asymmetry. Then they look at impulse responses from this split to output, finding that stimulation of output by inflation is markedly big for large differences between the expected and output neutral inflations and that this increases for high inflation.

VARman