بانک مرکزی و دلالان در بازار ارز خارجی تحت اطلاعات نامتقارن : یک رویکرد استراتژیک و شواهد

This paper studies the central bank sterilized intervention in the foreign exchange market in an asymmetric information framework. The central bank has inside information of its target, and the speculators have private information of fundamentals. The model suggests that if market expectations are precise about the central bank's target, or about liquidity trades, the foreign exchange market will more likely move in the direction opposite to the central bank's intervention direction. Following the hypothesis of Diebold and Nerlove [Diebold, F. X., & Nerlove, M. (1989). The dynamics of exchange rate volatility: A multivariate latent factor Arch model. Journal of Applied Econometrics, 4, 1–21] regarding the relationship between preciseness of information and the market volatility, the model provides a possible theoretical explanation for Hung's [Hung, J. H. (1997). Intervention strategies and exchange rate volatility: A noise trading perspective. Journal of International Money and Finance, 16, 779–793] noise trading channel hypothesis. Empirically, the paper shows that actual daily intervention data of the Fed, the Bundesbank, and the BOJ generally support the theoretical implications.

The central bank's sterilized intervention in foreign exchange markets draws strong interests not only from policy-makers and those from the academic circles, but also from investors. In the late 1990s, along with the actual daily intervention data being released by several central banks,1 empirical work has tended to use these data sets to test the actual effect of a central bank's intervention. Some find that sterilized intervention systematically affects the exchange rate in the short-run (Fatum & Hutchison, 2003); others find that the intervention exerts an incorrectly signed effect on the levels of exchange rates and tends to increase exchange rate volatility in the short-run (Beine, Benassy-Quere, & Lecourt, 2002). The literature also shows that the traditional arguments for sterilized intervention – the portfolio-balance channel and the signaling channel (the two major theories underlie the effectiveness of a central bank's sterilized intervention) – are weak, and the results are mixed.2 The disputes and mixed results call for further research. Bhattacharya and Weller (1997) look at this issue from a totally different perspective. Their paper studies the strategic aspects of a central bank's intervention by exploiting an asymmetric information model of sterilized intervention. The central bank has ‘inside information’ about its exchange rate target and trades with risk-averse speculators who have private information about future spot rates. The existence of private information means that an informed trader in the foreign exchange market will have an incentive to act strategically.3 Therefore, based on the strategic behavior of informed and uninformed traders, the model provides the circumstances where the central bank buying a currency is associated with depreciation of that currency (named “perverse response”) or with appreciation of that currency (named “normal response”). Furthermore, the model provides explanations for a puzzling issue in the intervention literature: the secrecy of central bank's intervention. It shows that secrecy about the scale of an intervention operation is always desirable, and secrecy about the target is sometimes desirable. This paper extends the work of Bhattacharya and Weller (1997) by modifying their model in two respects. First, Bhattacharya and Weller (1997) motivations for intervention are questioned based on results from a recent survey, available in Neely (2000).4 In Bhattacharya and Weller (1997), the profit made from forward transactions is part of the objective function of the central bank (together with limiting the variability of the exchange rate around target value T). However, the new survey shows that the profitability of intervention, as a motivation, is uniformly rejected by all central banks responding to the survey. Instead, two major motivations of central banks, ‘resist short-run trends’ and ‘seek to return exchange rates to fundamental values’, are supported by the survey results. This paper explicitly models these two motivations of the central bank. In this respect, the model, compared with that of Bhattacharya and Weller (1997), will be more realistic. The second improvement in the paper is that a noise term is included in the market clearing condition, which makes our model similar to the model introduced by Grossman and Stiglitz (1980). 5 The noise term may exist due to the random reduction/increase in demand for currency caused by ‘liquidity’ or reasons other than favorable private information. Bhattacharya and Weller (1997) argue that including the noise term in the market clearing condition makes their model intractable. However, by ignoring the noise term, at equilibrium, their model suffers a drawback—the speculators’ private information is fully revealed to the central bank, simply by observing the equilibrium price, and the central bank's target information is also fully revealed to the speculators, which will cause a unidirectional relationship between the central bank's intervention and the spot exchange. 6 This paper solves this problem by including the noise term originating from liquidity trades and argues that the randomness of the liquidity demand shock will dominate the central bank's intervention occasionally, and therefore break the unidirectional relationship. The major contributions of the paper are two fold. Theoretically, the paper develops a superior model by introducing the noise trading term into the equilibrium. Our model implies that the central bank should keep its intervention operations concealed, which is different from Bhattacharya and Weller (1997) in which they conclude that under certain conditions the monetary authorities prefer to reveal their objectives. Our conclusion is consistent with the implications of Vitale (1999) that the goals of sterilized intervention should never be disclosed. Two propositions are derived drawn from the model, namely: if the speculators have a high (low) precision of the central bank's target information View the MathML source(τTS), or the speculators have a high (low) precision of the volume of liquidity trader (τ x), then the central bank buying a currency is more likely associated with depreciation (appreciation) of that currency. Further, the two propositions are linked to market volatility following the hypothesis of Diebold and Nerlove (1989), which says that if the signals (information) are easily and unambiguously interpretable, then exchange rate volatility is likely to be low; if there is greater disagreement about the meaning of incoming information, or clearly relevant and significant information is scarce, the exchange rate volatility is likely to be high. Therefore, it is intuitively attractive to assume that, ceteris paribus, when the market volatility is low (high) before intervention, the speculators will have a high (low) precision of the central bank's target information View the MathML source(τTS), or the speculators will have a high (low) precision of the liquidity trade (τx). Thus, this paper provides a possible theoretic explanation for Hung (1997) noise trading channel hypothesis. According to Hung (1997) hypothesis, intervention is more likely to successfully change the direction of the exchange rate movement when the market volatility is high. One possible reason implied by this paper is that the speculators might have low precision of the central bank's target information or the liquidity trade when market volatility is high. Therefore, the central bank buying a currency is more likely to break the pattern of the currency movement. This paper also provides an empirical test of the theory. Based on the above assumptions, we examine the impacts of market volatility on the effect of central bank interventions. The market volatility is measured by a GARCH (1, 1) model. The results obtained from the Probit estimation unanimously support the implication that higher market volatility leads a lower probability of perverse response, therefore provide indirect empirical support for the two propositions. In addition, this paper finds that continuous interventions are more likely able to change the momentum of exchange rate movement. But, if the current exchange rate momentum is strong, it is much harder for the authorities to break the pattern. The size of the central bank intervention is found to have no systematic impact on the exchange rate trend. This finding implies that the signaling channel and noise trading channel (Hung, 1997) might be the most important two channels through which the interventions work, and that the portfolio channel is less relevant. One final empirical finding is that joint intervention seems to lead to a high perverse response, which is a surprise. However, it can be explained by the finding of Beine and Lecout (2004) that joint interventions are more transparent than otherwise. So this result is also consistent with the asymmetric information model's predictions. The plan of the paper is as follows. Section 2 provides the basic set up of the model. Sections 3 and 4 derives the optimal choice of the central bank and the speculators, respectively. Section 5 characterizes the equilibrium. First, Section 5 studies the demand function of the speculators at the equilibrium. Then, the spot exchange rate's response towards the central bank's intervention is derived. Based on the testable implications from the model, Section 6 provides empirical evidences supporting those implications. Section 7 provides some brief concluding remarks.

This paper extends the work of Bhattacharya and Weller (1997). The objectives of the central bank are the following two motivations: resisting short-run trends and seeking to return exchange rate to fundamentals values. A noise term is also introduced in the market clearing condition. One can assume that the noise term is caused by ‘liquidity’ demand. The modification implies that the central bank cannot fully uncover the private information by observing the equilibrium price, and the speculators cannot fully observe the central bank's target information. With the above changes, a 2, the term which determines the sign of exchange rate movement after the intervention, is found to be a linear function of the speculators’ precision of target View the MathML sourceτTS, which is different from the solution in Bhattacharya and Weller (1997). Our analysis is also different from Bhattacharya and Weller (1997) by suggesting that the goals of central bank intervention should never be disclosed, which is consistent with Vitale (1999) conclusions. Accordingly, if market expectations are precise about the central bank's target, or the volume of liquidity trades, ‘perverse’ responses of the foreign exchange market are more likely—central bank support intervention will tend to accompany a depreciation of that currency. One major contribution of this paper is that it provides a theoretical explanation of Hung (1997) hypothesis that a central bank is more likely to successfully change the direction of the exchange rate movement when market volatility is high. To show this, two propositions are derived from the above theoretic model, namely, if the speculators have a high (low) precision of the central bank's target information View the MathML source(τTS), or a high (low) precision of the liquidity traders (τ x), the equilibrium price of exchange rate will tend to have a perverse (positive) response to the central bank's intervention. Based on the hypothesis of Diebold and Nerlove (1989), the paper links the two propositions with the market volatility. According to Diebold and Nerlove (1989), if the signals (information) are easily and unambiguously interpretable, then exchange rate volatility is likely to be low; if there is greater disagreement about the meaning of incoming information, or clearly relevant and significant information is scarce, the exchange rate volatility is likely to be high. Therefore, it is reasonable to assume that, ceteris paribus, when the market volatility is low (high) before intervention, the speculators will have a high (low) precision of the central bank's target information View the MathML source(τTS), or the speculators will have a high (low) precision of the liquidity traders (τx). To test these two propositions, an empirical analysis is conducted using the actual daily intervention data of the Fed, the Bundesbank and the BOJ. Evidence show that a higher conditional market volatility is found to lead to a lower probability of perverse response (only the BOJ shows a weak significance level with p = 0.108), and more chances of switching the exchange rate trend for all three central banks. Therefore, the paper provides not only a theoretical explanation of Hung (1997) hypothesis that it is more likely to successfully change the direction of the exchange rate movement when the market volatility is high, but also empirical evidence. In addition, this paper finds that the continuous interventions are more likely able to change the momentum of exchange rate movement. But, as shown by the significant positive coefficient of ERDEVt, if the current exchange rate momentum is too strong, it is much harder for the authorities to break the pattern. One surprising finding of this paper is that joint interventions seem to more often lead to a perverse response. This may be due to two facts: on the one hand, the proportion of secret interventions is found to be lower for concerted operations (Beine & Lecout, 2004); on the other hand, it is harder for the authority to hide its hand while conducting joint intervention. According to the theory, a higher proportion of perverse responses are not a surprise. Finally, the size of the central bank interventions is found to have no systematic impact on the exchange rate trend. This finding implies that the signaling channel and the noise trading channel of Hung (1997) might be the most important two channels through which the interventions work; the portfolio channel is less relevant.