Nonlinearity and asymmetry in the monetary policy reaction function: a partially generalized ordered probit approach

Abstract

The present paper employs a partially generalized ordered probit method to model the Federal Reserve’s monetary policy reaction function. The partially generalized ordered probit method eliminates the parallel regression assumption (which is assumed in ordered probit models) and reveals an important new asymmetry in the Federal Reserve’s actions. The findings indicate that a general monetary reaction function outperforms standard Taylor rule specifications. The Fed takes into account not only inflation and output measures but also several other variables during its decision process, but the degree of its attention on each variable is choice-dependent. The Fed might assign different weights for each macroeconomic factor when it is trying to make a choice, for example, between a big and small decrease or a small decrease and no change in the federal funds target rate. The threshold estimates also indicate that the Federal Reserve acts asymmetrically that it waits for relatively significant changes in the macroeconomic factors before it decides for a change in its target rates. However, once these thresholds are passed, relatively less significant changes in the economy are needed for the Federal Reserve to take action.