Target zone models postulate that `without intervention' exchange rates are driven by their fundamentals, while policy-makers' obligations to intervene in foreign exchange markets as exchange rates approach the margins of the band give rise to speculation that stabilizes exchange rates within the band. Although this typical `target zone' effect has several testable empirical implications, little empirical support for such effects has appeared to date. Conversely, many of the empirical regularities of exchange rates observed in managed exchange rate arrangements, such as the Exchange Rate Mechanism (ERM) of the EMS, have been found to be largely inconsistent with the predictions of the simple target zone model.

In Discussion Paper No. 554, Research Fellow Axel Weber develops a new approach to the empirical evaluation of target zone models, based on the hypothesis that observable exchange rates in target zones whether perfectly or imperfectly credible must frequently exhibit stochastic process switching, which may be caused by stabilizing intervention, stabilizing speculation or realignments. Using a Bayesian approach to estimate the probability of such stochastic process switching, Weber finds that this was an important feature of the EMS in its early period, as predicted by exchange rate target zone models. Such frequent switching has recently vanished, however, which suggests that the EMS has now converged to a system of `credible target zones', in which exchange rates fluctuate in an unregulated random walk but do not permanently drift outside the bands. Weber attributes this result to the convergence of EMS members' inflation and interest rates to the low `German' levels, which may provide the necessary conditions for the proposed tightening of target zones during Stage II of the Delors Plan. The absence of stochastic process switching suggests that the current bands of 2.25% around parity may be too wide to induce stabilizing speculation, so moving ahead with Stage II may revive the potential merits of currency bands.

Weber also finds that the estimated probabilities of intervention derived from the actual behaviour of EMS exchange rates vis- à-vis the Deutschmark closely resemble the time-paths of EMS interventions by the Bundesbank; so a non-zero probability of stationary exchange rate movements typically coincides with stabilizing Bundesbank intervention, which generally takes the form of obligatory intervention at the boundaries of the band. Weber concludes that these two results together leave little scope for attributing stochastic process switching in the EMS currencies' exchange rates vis-à-vis the Deutschmark to stabilizing speculation of the type postulated in target zone models.

Stochastic Process Switching and Intervention in Exchange Rate Target Zones: Empirical Evidence from the EMS
Axel Weber

Discussion Paper No. 554, July 1991 (IM)