Testing Target Zone Credibility with a Limited Dependent RE Model

Meral Karasulu
Boĝazici University, Istanbul
Karasulu@boun.edu.tr

Abstract

This essay applies a limited dependent variable rational expectations model to estimate the French Franc/Deutsche Mark (FF/DM) exchange rate in order to test the credibility of the European Exchange Rate Mechanism (ERM) for these currencies.

The existing literature treats the exchange rate within the band as an unbounded continuous variable. The exchange rate within a target zone is a bounded variable censored beyond the upper and lower band margins. If the censored nature of the exchange rate is ignored the parameter estimates will be biased. The confidence intervals for expected realignment will reflect this bias and distort the result of credibility tests.

The limited dependent variable rational expectations model integrates this feature of the exchange rate mechanism into the estimation process. Moreover, it explicitly models expectations of economic agents who will incorporate the bounded nature of the data into their information sets when forming expectations about exchange rate movements.

I find considerable support for modeling the exchange rate using limited dependent variable framework. My results indicate that accounting for the band in the drift adjustment method significantly modifies the results of credibility tests. As opposed to earlier findings in the literature, confidence intervals for the expected realignment are wider than those resulting from ordinary least squares estimates and from a rational expectations model obtained without imposing the band. Furthermore, my results trace the history of the ERM realignments for the FF/DM and in each case correctly identify the period before and after a realignment as a period lacking credibility. The model predicts a gradual widening of confidence intervals after 1980 until the end of 1990. Although a zero expected realignment is contained in the confidence intervals for most of this period, the widening of the interval also suggests a tension building up with possibly larger realignment sizes.


Society of Computational Economics
Second International Conference on Computing in Economics and Finance
Geneva, Switzerland, 26-28 June 1996