تجارت آموزنده یا فقط سر و صدای پرهزینه؟ تجزیه و تحلیل مداخلات بانک مرکزی

We conduct an empirical investigation of the impact of Central Bank intervention on the process of price formation in foreign exchange markets. The main contributions of this paper are (i) in considering the effects of official interventions on multiple dimensions of such a process beyond the first and second moment of currency returns and (ii) in exploiting insights from the analysis of market liquidity in proximity of these trades to explain their effectiveness. For that purpose, we employ a unique dataset of tick-by-tick indicative quotes and intraday (informative) sterilized spot interventions and (uninformative) customer transactions executed by the Swiss National Bank (SNB) on the Swiss Franc/U.S. Dollar exchange rate (CHFUSD) between 1996 and 1998. Using several empirical strategies (some of which are novel to the exchange rate literature), we find that the effectiveness of these trades is crucially related to their perceived information content, rather than to imperfect substitutability or inventory considerations. Indeed, regardless of their size, only SNB interventions (especially when unexpected or leaning against the wind) have significant and persistent effects on daily CHFUSD returns, although they often fail to smooth currency fluctuations. Nonetheless, only SNB interventions, regardless of their effectiveness, induce significant misinformation and heterogeneity of beliefs among market participants and deteriorate market liquidity. These externalities always translate into higher, economically significant transaction costs borne by investors.

Central Bank interventions are one of the most interesting and puzzling features of the global foreign exchange (forex) markets. Central Banks often engage in individual or coordinated efforts to influence exchange rate dynamics, either to strengthen or resist market momentum, to calm disorderly market conditions, to signal current or future economic policies, or to replenish previously depleted reserves. There is strong consensus in the economic literature (e.g., Adams and Henderson, 1983) that unsterilized interventions, by affecting the existing stock of high-powered money, influence the exchange rate through the traditional channels of monetary policy. The effectiveness and necessity of sterilized interventions, i.e., those accompanied by offsetting actions on the domestic monetary base, are instead still controversial, and, as such, at the center of the current theoretical and empirical debate. 1 Within the standard macroeconomic approach, sterilized interventions may affect the exchange rate through either of two channels: imperfect substitutability or signaling. The first channel is usually examined in the context of portfolio balance models (e.g., Branson, 1983 and Branson, 1984) in which risk-averse market participants need to be compensated for holding sub-optimal portfolios following the intervention. The second channel (Mussa, 1981 and Bhattacharya and Weller, 1997) allows sterilized intervention to affect prices by conveying not only information on policy intentions, but also fundamental information about the future value of the currency. Yet, according to Dominguez (2006, p. 1052), “neither of these channels is easily reconciled with the empirical evidence” on whether and how interventions influence exchange rate movements and volatility, especially in the short run.2 Within the newer market microstructure approach to currency determination (see Lyons, 2001 and Evans and Lyons, 2002), theoretical research concentrates on the process through which traders revise their expectations and dealers adjust prices, either temporarily or permanently, in response to sterilized interventions (e.g., Evans and Lyons, 2003, Vitale, 2003 and Pasquariello, 2005). Two recent empirical advances, surveyed in depth by Neely (2005), have enhanced our understanding of these mechanisms. The first is the availability of high frequency data on both exchange rates and Central Bank interventions. The second is the use of event studies to examine the behavior of exchange rates in proximity of intervention dates or trades with minimal assumptions about the data-generating process. Along these lines, many studies concentrate on the impact of these transactions on daily or intraday exchange rate returns and return volatility.3 Despite its promise and encouraging results, this research also suffers from several limitations. High-frequency exchange rate data are in fact plagued by microstructure frictions, many of which are potentially relevant over intraday event intervals. Further, both interventions and exchange rate fluctuations are likely to be determined simultaneously, since Central Banks respond to undesired market conditions. Endogeneity biases may then affect both significance and interpretation of estimates of the contemporaneous reaction of the exchange rate to Central Bank transactions. The analysis of intraday event windows around precise event times can only attenuate, but does not eliminate this concern, since interventions may display their full effects over days or weeks.4 Finally, most event studies examine the impact of interventions on specific facets of the forex market in isolation, hence ignoring both their interaction and the additional information they may provide to evaluate the effectiveness of these policies. Instead, the presence of active price manipulators in the otherwise very liquid and (widely recognized as) efficient forex markets raises a broader question: what is the impact of interventions on the process of price formation in the currency market as a whole, hence not only on exchange rate dynamics but also on the market's ability to process information and investors’ ability to trade? This question, albeit of interest to investors, analysts, and policymakers, has so far received little attention.5 Addressing this question is the main contribution of our study. We do so by bridging both the macro and microstructure literature on Central Bank intervention. Specifically, we investigate sterilized spot interventions and passive trades (commonly labeled “customer transactions”) executed by the Swiss National Bank (SNB)—the Swiss Central Bank—on the Swiss Franc/U.S. Dollar exchange rate (CHFUSD) between 1986 and 1998. The CHFUSD is among the most liquid currency pairs traded in the global forex markets, the SNB is one of the most credible monetary authorities in the financial world, and its interventions are contemporaneously publicly available. Although SNB interventions may be informative about economic fundamentals and policy motives, customer transactions are not, since they are conducted by the SNB for reasons other than exchange rate management. We then build a database matching those transactions with tick-by-tick quotes posted by dealers on Reuters terminals and use them to compute daily exchange rate returns and measures of ex post volatility, transaction costs, and trading intensity. By studying the lower-frequency impact of intraday SNB trades on daily aggregates, we focus on longer horizons—of relevance to policymakers—without discarding valuable intraday information, while minimizing potential distortions to the inference from microstructure frictions. 6 Further, inference from the estimated impact of SNB trades on market liquidity is less likely to be biased by endogeneity considerations, since transaction costs and trading intensity are less likely to enter the reaction function of a Central Bank in a significant way. However, insights from the analysis of market liquidity in proximity of its interventions can be exploited to learn about the nature of their effectiveness. Our analysis proceeds in two stages. In the first stage, to motivate our effort, we use an event study methodology to examine the cumulative impact of SNB trades on each of those daily aggregates separately over an interval of roughly three weeks around the event dates. We find that SNB interventions, despite accounting for only a small portion of the average daily turnover in the CHFUSD market, significantly affect these variables both in the short and long run. Regardless of their size, SNB interventions are unforeseen (except those chasing the trend) and have persistent effects on currency returns (lasting for several days after their execution), especially when against existing momentum. The SNB is much less successful in calming disorderly market conditions: ex post measures of currency volatility always surge in proximity of its trades and stay high for many days afterward. These trades are nonetheless costly: absolute and proportional spreads often widen in their proximity, and market liquidity deteriorates. The surge in spreads is economically significant as well: we estimate that (annualized) transaction costs borne by investors and speculators increase by almost $150$150 million when the SNB sells USD and by over $815$815 million when the SNB buys USD. These results are robust to several extensions and variants of our basic empirical strategy, as well as to the inclusion of potentially important economic variables for the decision-making process of both the SNB and the Federal Reserve. In the second stage, we assess the relative importance of inventory, risk-aversion, adverse selection, and information considerations in explaining those non-trivial changes in exchange rate returns, return volatility, and market liquidity. For that purpose, we extend the existing literature in three directions. First, we repeat our preliminary analysis for a control sample made of all customer transactions executed by the SNB over the sample period. Indeed, since these transactions are by their nature uninformative, they provide a unique benchmark to gauge the relevance of the information content of SNB trades on the effects described above. We find that those transactions have a negligible impact on the many dimensions of the microstructure of the CHFUSD market.7 Second, we estimate a series of bivariate vector autoregression (VAR) models of auto- and cross-correlations between interventions and either quote revisions, spreads, or trading intensity (i) to account for their potentially simultaneous determination and (ii) to identify the unexpected component of each SNB trade, without making explicit assumptions on the structure of their interaction. Within this setting, any permanent impact of these trades on quotes or market liquidity can be attributed solely to its surprise information content, rather than to transient inventory effects (Hasbrouck, 1988 and Hasbrouck, 1991). The ensuing evidence indicates that, when unanticipated, SNB interventions have an even more pronounced directional effect on the Swiss Franc, again at the cost of even greater volatility and transaction costs borne by investors. Third, we focus explicitly on shocks to transaction costs in proximity of SNB interventions. Specifically, we decompose daily bid–ask spread shocks during those trades into shocks related to misinformation, liquidity, fundamental volatility, competition, and immediacy using a reduced-form model (adapted from Fedenia and Grammatikos, 1992) for the many theories of spread determination in the market microstructure literature. We then use the relative significance of these arguments to interpret the impact of SNB interventions on CHFUSD returns and return volatility. Estimation of the model reveals that information and liquidity shocks, but not portfolio balance effects, explain the increase in transaction costs typically accompanying SNB transactions. In particular, especially when large and chasing an existing trend, these trades induce the greatest heterogeneity of beliefs among market participants, which translates into a small impact on currency returns and wider spreads.8 Overall, our empirical analysis suggests that Central Bank interventions have important effects on the many dimensions of the process of price formation in the currency markets, and that information (rather than inventory or portfolio balance) considerations are the single most important determinant of these effects. Nonetheless, regardless of their effectiveness in influencing the dynamics of the target exchange rate, official interventions appear to cause severe externalities—often in the form of deteriorating market liquidity—that cannot be exclusively attributed to adverse selection. Hence, Central Banks may face an economically significant trade-off between (mis)information and transaction costs when formulating and executing their currency management policies. The paper is organized as follows. Section 2 describes our dataset. In Section 3 we analyze the reaction of currency returns, return volatility, and market liquidity to SNB transactions in isolation, explore the relevance of important attributes of these trades (size, momentum, expectations) for our findings, and perform several robustness checks. The decomposition of estimated spread shocks during SNB interventions is in Section 4. Section 5 concludes

Are official interventions in the forex market a source of information or just noise? Is there any cost borne by investors stemming from Central Banks’ attempts at managing currency fluctuations? And if so there is, why? Providing an answer to these questions by analyzing both the impact of Central Bank interventions on the process of price formation in the forex markets and the interaction among its many dimensions constitutes the contribution of this study to the exchange rate literature. Our analysis shows that sterilized SNB interventions considerably affect different measures of exchange rate behavior, ex post volatility, market liquidity, and trading intensity both in the short and in the long run. Using an event study methodology (as well as alternative empirical strategies) on a joint dataset of indicative quotes and SNB transactions, we find that these signed trades, although representing only a small fraction of the average daily turnover in the CHFUSD market, have meaningful, asymmetric, and persistent effects on currency returns, especially when leaning against the wind, regardless of their size. Interestingly, the Swiss Franc market does not anticipate incoming interventions except when chasing the trend. Lastly, official USD purchases tend to follow a steady strengthening of the CHF, while official USD sales frequently come in reaction to a period of protracted CHF weakness, consistent with historical accounts of the monetary policy activity of the SNB (Rich, 1997 and Peytrignet, 1999). The SNB is much less successful in smoothing fluctuations of the currency or in reducing its variability, at least in the short run. Ex post measures of exchange rate volatility in fact always surge in proximity of interventions and stay high for many days afterward. This choice may, however, be optimal: for example, Rich (1997) argues that short-term currency volatility may allow the Swiss Central Bank to respond to unexpected shifts in preferences, hence to preserve price stability in the long run. Yet, despite their effectiveness, SNB interventions are also costly to investors: absolute and proportional spreads for the CHFUSD widen in a (statistically and economically) significant fashion around those trades, often prior to the actual intervention event. Because these effects are more pronounced in response to unanticipated SNB interventions, but are negligible in the control sample made of ex post uninformative customer transactions, we conclude that the potential information content of SNB interventions must play an important role in explaining their influence on the CHFUSD market. Consistently, the decomposition of spread shocks in proximity of SNB trades further reveals that a significant portion of the increase in transaction costs can be explained by greater information heterogeneity and fundamental volatility and lower market liquidity and competition among dealers. In particular, large and trend-chasing interventions, the least successful in our sample, induce the greatest misinformation across market participants and reduce trading immediacy. These findings have meaningful policy implications. Indeed, our analysis indicates that official interventions, even when effective against market momentum or disorderly market conditions, are often costly as well, not only for the Central Bank but also for investors and speculators. The evidence reported in this study suggests that this trade-off is complex, persistent, and non-trivial. It should therefore be at the center of any currency policy debate