In this paper, we analyse the effectiveness of the direct central bank interventions using a new effectiveness criterion. To this aim, we investigate the effects of central bank interventions (CBI) in a noise trading model with chartists and fundamentalists. We first estimate a model in which chartists extrapolate past returns and fundamentalists forecast a mean reverting dynamics of the exchange rate towards a fundamental value. Then, we investigate the role of central bank interventions for explaining the switching properties between the two types of agents. We find evidence that in the medium run, interventions increase the proportion of fundamentalists and therefore exert some stabilizing influence on the exchange rate.

Intervening in the FX market implies a complex decision process for central banks. Monetary authorities have to decide whether to intervene or not, and if so, when and how. Since the successive steps of this procedure are likely to be highly interdependent, we adopt a nested logit approach to capture their relationships and to characterize the prominent features of the various steps of the intervention decision process. Our estimations based on Japanese data from 1991 to 2004 indicate that the Bank of Japan: (i) mainly reacted to deviations of the exchange rate with respect to fundamentals and (ii) tended to favour secrecy when its credibility was low. We also provide new insights on the so-called secrecy puzzle by modeling explicitly the risk for a secret intervention to be detected. Our results have important implications in terms of exchange rate policy, such as the emergence of a trade-off between intervention size, communication policy and secrecy. Our results tend to provide some explanation for the observed persistence of ineffective intervention policy during some sub-periods.