Central bank intervention and market expectations

BIS Papers  |  No 10  | 
01 April 2002

In this paper we illustrate a new analytical method and present new results on the relationship between foreign exchange market intervention and market expectations in the dollar/mark and dollar/yen markets between 1985 and 1996, and 1991 and 1996, respectively. The paper improves on previous work in several important respects. First, we use official data on intervention carried out by the G10 central banks in the dollar/mark and dollar/yen markets from 1985 to 1996 and from 1991 to 1996, respectively. Second, we also consider data on market expectations of intervention derived from press reports to investigate the extent to which markets were driven by perceptions of intervention as well as actual intervention. Third, we broaden the perspective on the relationship between intervention and market expectations by looking at the entire expected distribution of future exchange rates. The four moments of estimated probability density functions (PDFs) allow a more complete characterisation of the state of market expectations on a particular day. Fourth, we use a data set on macroeconomic variables that allows for the influence of factors other than intervention on exchange rates.

We follow two complementary approaches to analyse the relationship between intervention and market expectations. The first is based on an event analysis, and investigates how the moments of the PDFs changed around a number of important intervention episodes. The main advantage of this approach is that it can identify the context in which each particular intervention episode occurred and the objectives that central banks were actually pursuing. The second approach looks at averages over episodes and uses econometric techniques to reveal broad average tendencies. It has the advantage that it can summarise results for a large number of individual episodes, while at the same time controlling for the influence of factors other than intervention.

Based on the event study methodology, we conclude that, depending on circumstances, particular interventions did succeed in affecting traders' expectations of future exchange rate movements in line with policymakers' objectives. We also find that the impact of interventions varies considerably across episodes. By contrast, when using econometric techniques to look at the average experience for different sub-periods, we find no evidence that intervention on its own had a statistically significant, systematic impact on expected future exchange rates, where statistical significance is measured at the usual significance levels. Likewise, while there is some evidence that for the period 1992-96 concerted interventions may have had a stronger impact on market expectations, econometric results suggest that different intervention strategies did not seem to have systematically dissimilar effects at the 95% or 90% confidence level. Hence, the interpretation of our results depends in part on the reader's views on the two approaches that we use, as well as the ultimate objectives of intervention in particular circumstances.