Testing credibility with time-varying coefficients

Abstract

The article presents evidence on credibility gains from the adoption of Inflation Targeting (IT), for a set of developed and emerging markets. We carry out our analysis for Brazil, Mexico, Sweden and the UK, and ask whether the adoption of this monetary framework strengths monetary policy credibility. To answer this question we apply a time-varying coefficients methodology to estimate the time path of the trade-off between inflation and output. Following the literature a more favourable trade-off is believed to be linked to improvements in monetary policy credibility. The results show a clear picture of credibility gains from the adoption of IT.

Notes

¹ Eichengreen (2002) observed that emerging markets trying to enhance the credibility of their monetary policy tend to provide higher levels of openness and transparency than developed economies.

² Examples of the use of the Kalman Filter in monetary policy credibility studies that take a state-space form can be found in Hardouvelis and Barnhart (1989) and Agenor and Taylor (1992, 1993).

³ As discussed by Andersen and Wascher (1999) the model also assumes that the aggregate demand curve is unit elastic and that shifts in the aggregate supply and demand curves are independent from each other. Although these assumptions may not be valid, they also apply to most of the models of the inflation-output trade-off.

⁴ Two advantages of Schlicht (1989) methodology in comparison to the Kalman Filter are: one, it is double-sided, i.e. it uses all the information available to compute the coefficients, whereas the Kalman Filter is one-sided, i.e. uses only past information; and two, Schlicht (1989) estimator uses an orthogonal parameterization instead of the usual parameterization by initial values, which effaces the problems associated with estimating, or otherwise providing, initial values. For a comparison between simulations using the Kalman Filter and Schlicht (1989) methodology, see Schlicht and Ludsteck (2006).

⁵ Note that when estimating the model an intercept was included to Equation 2(2).

⁶ Equations 2(2) and 3(3) are not estimated together, as Schlicht's VC software for time-varying estimations only allows the estimation of one equation at a time.

⁷ The results are not reported here but are available on request.

⁸ The Real Plan was put in place in 1994:07, and was very successful in bringing down inflation in Brazil to lower and declining rates.