مطالعه میزان معامله اثرات مداخله بانک مرکزی در نرخ ارز

We study the effects of sterilised intervention operations executed on behalf of the Swiss National Bank (SNB) using tick-by-tick transactions data between 1986 and 1995. We extend the preliminary analysis of [Economic Journal 109 (1999) 662] by matching these data with indicative intra-day exchange rate quotes and news-wire reports of central bank activity. Using an event study approach we find that intervention has important short-run effects on exchange rate returns. In particular, among various results, we find that i) intervention has a stronger impact when the SNB moves with-the-market and when its activity is concerted with that of other central banks and ii) exchange rate returns move in the 15 min interval prior to interventions.

Scholars have long debated whether central bank intervention operations in the markets for foreign exchange have important effects on exchange rate levels and volatility and on market conditions. While it can be theoretically established that sterilised intervention affects the value of currencies and level of activity in FX markets, either through a portfolio-balance effect or via a signalling channel (Mussa, 1981, Bhattacharya and Weller, 1997, Vitale, 1999 and Vitale, forthcoming), the effectiveness of sterilised intervention remains an unresolved issue from an empirical point of view. A contributory factor to the unresolved nature of this issue has been a lack of adequate data on central bank intervention transactions. Indeed, until recently, researchers only had access to data sets in which intervention operations were aggregated to daily or lower frequencies.1 This has proved a serious impediment to empirical analysis of the effects of intervention on exchange rates as, with coarsely sampled data, it is difficult/impossible to overcome simultaneity problems and to characterise the high-frequency effects of intervention on market conditions. However, recently the Swiss National Bank (SNB) has made an innovative data set available to researchers, containing tick-by-tick observations on its intervention operations in FX markets between 1986 and 1995. We extend a preliminary analysis of this data set by Fischer and Zurlinden (1999), by combining the information it contains with indicative exchange rate quotes recorded by Olsen and Associates and with Reuters news-wire reports of central banks’ activity. These data allow us to conduct a high-frequency analysis of the effects of signed intervention operations on exchange returns. We construct time-series for the USD/CHF exchange rate and signed intervention quantities sampled once every 15 min and use these data to analyse the effects of intervention via an event study.2 The particular focus of our study is the empirical relevance of the signalling hypothesis. This hypothesis suggests that intervention operations are used by monetary authorities to convey information to FX markets and hence alter market expectations and exchange rates. Thus, if central bank operations are informative signed intervention should have a significant and permanent effect on the value of currencies. Other important issues we can analyse using the event study methodology include: i) the speed with which intervention influences markets, ii) the impact of intervention size and iii) the effects of market conditions on the effectiveness of intervention. With respect to the first issue, it is usually presumed that FX markets are very resilient and process information very quickly. The current analysis gives us an opportunity to test this assertion. Analysis of the impact of intervention size is also important, in that the signalling hypothesis suggests that intervention is effective because it is potentially expensive. Hence larger trades should have a bigger impact on exchange rates and market characteristics. Finally, we investigate whether interventions that seek to reinforce prior exchange rate movements have different effects than those which seek to reverse such movements. The main findings from our analysis are the following. 1. Intervention events have a positive effect on exchange rates: when the SNB purchases (sells) US dollars, the American currency appreciates (depreciates). The impact of intervention is immediate, as the exchange rate moves within the 15 min interval during which an intervention event is reported, and persistent, as the cumulative effect is still significant after few hours. We also observe market anticipation of intervention information, as the exchange rate moves (significantly) in the direction of the operation in the 15 min interval that precedes that in which the event occurs. Likewise, we detect partial reversal of the intervention effect in the 15 min period immediately after the event interval. 2. These conclusions hold: i) when we simply consider the direction of intervention as an explanatory variable, ii) when we employ the signed intervention size in the regression analysis of the exchange rate return and iii) even when we account for the effects of the intervention activity of other central banks, notably the Fed and the Buba. On the contrary these conclusions are not valid when we use non-intervention trades carried out by the SNB instead of intervention operations. 3. The impact of SNB activity is larger and more persistent: i) when operations are concerted with those of the Buba and the Fed than when they are conducted unilaterally and ii) when intervention follows the current trend rather than opposing it. The rest of the paper is organised as follows. In Section 2 we briefly describe our data set and the statistical properties of the exchange rate and intervention series. In Section 3 we present our results. Section 4 concludes.

We have conducted an investigation of the effects of sterilised intervention in the spot USD/CHF market. The novelty of this study lies in the use of a transaction based data set of SNB activity between 1986 and 1995, the information from which we combine with indicative exchange rate quotes recorded by Olsen and Associates and Reuters news-wire reports. With such a rich data set we have been able to identify a clear and significant link between FX intervention and exchange rate returns. In particular, using an event study, we have exactly quantified the effects of single intervention operations on the USD/CHF rate at a 15 min sampling frequency. Our analysis suggests four important results: i) SNB intervention operations have strong and persistent effects on the USD/CHF which remain after accounting for the operations of other central banks; ii) SNB interventions are more effective in conditioning exchange rates when they are coordinated with other central banks; iii) interventions which are with-the-trend have stronger exchange rate impacts; iv) the exchange rate moves in the direction of the intervention in the minutes before the actual intervention takes place. Clearly, the latter result is puzzling as it may indicate that some market participants learn about upcoming SNB interventions shortly before they actually occur. We find that the result persists even after accounting for the effects of co-ordinated intervention among central banks using news-wire reports of Fed and Buba activity. Our interpretation of this result is that the SNB chose to intervene with-the-wind on a very high-frequency basis i.e. it is not that intervention affects returns one-period ahead of its reported time (due, for example, to leakage of intervention information) but that intervention is timed by the SNB to follow high-frequency price moves in the direction that the SNB desires to push the exchange rate. Perhaps this is due to the SNB recognising that interventions that follow a recent trend are more likely to be successful than those opposing recent trends. A clearer picture of the anticipated intervention effect could be given via a structural analysis of the effects of intervention on exchange rates, as this would solve identification problems and isolate the effects of the unexpected components of intervention. However, such an analysis is very hard to implement, in that it is quite difficult to predict intervention operations. This difficulty is particularly acute for the initial operations within single intervention episodes. This is unfortunate, because exactly these operations appear ( Fischer and Zurlinden, 1999) to be responsible for most of the effects on exchange rates of FX intervention. We leave this avenue to future research.