مداخله بانک مرکزی و نرخ ارز خارجی: شواهد جدید از برآورد فیگارچ

In this paper, we investigate the effects of official interventions on the (short run) evolution and volatility of exchange rates. To this aim, we rely on a new measure of volatility implied by the FIGARCH model that outperforms the traditionally used GARCH one. It is found that central bank interventions exert an incorrectly signed effect on the levels of exchange rates and tend to increase their volatility in the short run. In general, our results also show that the traditional GARCH estimations tend to underestimate the effects in terms of volatility.

There is some extensive evidence showing that real exchange rates became more volatile when the nominal rates were allowed to float, and that such exchange rate volatility may have destabilized domestic economies (see Flood and Rose, 1995). Such evidence is part of the rationale for the historical attempts to reduce exchange rate volatility through European exchange rate arrangements and monetary union. It also justifies the proposals for stabilizing exchange rates within fundamentals—determined target zones (Williamson, 1985 and Williamson and Miller, 1987 and more recently, Bergsten and Miller, 1999). However, in the latter case, the reorganization of the International Monetary System can be successful only if central banks have powerful instruments to stabilize the exchange rates around what they think to be their fundamental values. The two available instruments are monetary policy and direct intervention in the foreign exchange market. These instruments are interrelated as far as official interventions are not sterilized. However, interventions can also have a signaling effect, i.e. they can provide information about future monetary policy (Dominguez and Frankel, 1993). In this case, the impact of interventions is independent from contemporaneous monetary policy. Existing evidence on the impact of official interventions on exchange rates is rather mixed. A first generation of studies provided estimations based on quarterly variations of official reserves1. They showed little impact of these variations on exchange rates, especially when interventions were sterilized. However, the proxies used for interventions were subject to valuation effects which have nothing to do with interventions. In addition, such studies based on quarterly data could not assess the short-run impact of interventions. Since the early 1990s, the question of the effectiveness of central bank interventions on the foreign exchange market has nevertheless received a renewed interest due to the public release of daily data of interventions by several major central banks over past periods, and to the development of econometric techniques for daily data displaying non-normal distributions as well as time-dependent conditional variance. A second generation of empirical work devoted increasing attention to the effects on volatility (Bonser-Neal and Tanner, 1996, Baillie and Osterberg, 1997a, Baillie and Osterberg, 1997b and Dominguez, 1998) which was the cornerstone of the 1987 Louvre Agreement. Quite surprisingly, the literature as a whole concludes that interventions either have no impact, or have a positive effect on exchange rate volatility. In addition, there is some moderate evidence that the purchase of dollars by central banks was associated with subsequent dollar depreciation, which is often interpreted as a “leaning-against-the-wind” behavior from the central banks (Baillie and Osterberg, 1997b)2. In this strand of literature, two approaches have been adopted to analyze the effects of central bank interventions on exchange rate volatility. The first approach focuses on exchange rate volatility expectation. For instance, a usual strategy makes use of measures of expected volatility derived from option prices (Bonser-Neal and Tanner, 1996 and Galati and Melick, 1999). The second approach relies on another measure of exchange rate volatility which is drawn from econometric models allowing the variance to change over time. These models belong to the well known ARCH and GARCH approaches initiated by Engle (1982) and Engle and Bollerslev (1986). The latter approach allows us to study the effect of interventions on both the levels and the volatility. Up to now, the above-mentioned analysis based on ex-post measures have exclusively relied on a GARCH specification. Indeed, the GARCH framework seemed to be a well-established model and thus a reliable starting point in analyzing volatility. Nevertheless, the relevance of the GARCH framework—and its special case, the Integrated GARCH model—has recently been questioned. Indeed, the stable GARCH model implies that the effect of a volatility shock vanishes over time at an exponential rate. By contrast, the IGARCH model implies a complete persistence of such a shock. These features stand in sharp opposition to stylised facts drawn from the study of financial markets. To cope with this issue, Baillie et al. (1996) introduce the Fractionally Integrated GARCH (FIGARCH) model that allows for some persistence of volatility shocks more in line with these facts. Empirical applications of this model to the major daily exchange rates (Baillie et al., 1996, Tse, 1998 and Beine et al., 2002) confirm the relevance of this new framework. The purpose of this paper is to assess the effects of central bank interventions on the variation and volatility of the Deutsche mark/US dollar (DEM and USD hereafter) and the Japanese YEN/USD exchange rates using the FIGARCH framework. We compare the results with those of the literature and henceforth assess the importance of relying on a more appropriate measure of volatility. Furthermore, we follow Dominguez (1998) as well as Bonser-Neal and Tanner (1996) in distinguishing the nature of central bank interventions (coordinated or not, reported or secret) but over a longer period (1985–1995). Our results only partly support the previous findings. In particular, it is shown that ignoring the finite persistence of volatility shocks can result in a spurious rejection of the effects of interventions on the volatility of exchange rates. Moreover, our results exhibit some homogeneity across all the major sub-periods of interventions, including the most recent one (1992–1995) that was not investigated in the previous literature. The paper is organized as follows. Section 2 recalls the technical background of the FIGARCH model. Section 3 presents the data. Section 4 tests for the effects of central bank interventions for the DEM/USD and YEN/USD. Section 5 proposes additional estimations aiming at assessing the robustness of the results and refining their economic interpretation. Section 6 concludes.

In this paper, we have investigated the empirical effects of the major central bank interventions on the short run dynamics of the major nominal exchange rates (DEM and YEN) in terms of the USD. To this aim, we have relied on a new econometric framework, the FIGARCH model that yields a more appropriate measure of the ex-post volatility of the exchange rates than the traditional GARCH approach does. Indeed, the FIGARCH model implies a more realistic dynamics of the persistence of volatility shocks and is strongly supported by the data over all the periods under investigation. As a whole, our FIGARCH estimations show that official interventions manage to move the market (especially when they are reported in the press), but in the wrong direction. Further investigations show that our results do not systematically stem from reverse causality, although central banks clearly lean against the wind. Consistently with the previous literature, central bank interventions are found to increase exchange rate volatility. Furthermore, through the FIGARCH specification, this positive impact is shown to display some persistence. To sum up, our findings point out the limited efficiency of central bank interventions as a stabilization instrument in the short run and emphasize the cost in terms of exchange rate volatility. Another important point lies in the different results associated to the GARCH and the FIGARCH frameworks. It is shown that the traditional GARCH estimations used in the literature tend to underestimate the (positive) effects of the central bank interventions on the ex-post volatility of exchange rates. In turn, this illustrates that measuring the volatility through the FIGARCH approach instead of the GARCH one matters in empirical work.