کارایی بازار در بانک مرکزی : مورد تایوان

This paper investigates the empirical relation between order imbalance and intraday NTD/USD exchange rate dynamics. Using one-year high frequency data, we demonstrate that interbank order imbalances have substantial explanatory power for concurrent exchange rate returns both on the daily and intraday bases. More importantly, we find that lagged-one order imbalances have a predictive negative effect on current returns. Further, we trace the predictability of returns using order imbalances over various intervals to investigate the intraday market efficiency. We show that the weak-form efficiency appears to prevail over intervals of 15 to 60 min in the NTD/USD exchange rate market.

This paper provides an empirical study of relation between order imbalance and intraday exchange rate dynamics in the case of a small open economy, Taiwan, with a managed floating exchange rate regime for local currency, the New Taiwan Dollar (NTD). Based mainly on the microstructure model of Evans and Lyons (2002a), we propose the time-series regression model and a time-varying GARCH(1,1) model to capture NTD/USD dynamics employing one-year inter-dealer high frequency data. We investigate the predictability of order imbalance on NTD/USD return by running time-series regressions of price movement on lagged imbalances as well as the out-of sample performance compared with random walk model. Furthermore, we trace the predictability of returns using order imbalances over various intervals to investigate the intraday efficiency. The choice of the exchange rate studied is essential for policy makers of small open economies for several reasons. First, the exchange rate is perhaps the most important asset price in the globalizing economy (Rose, 2011). A floating currency is a key component of monetary policy framework, helping the economy to adjust to shocks and playing an important part in the transmission of monetary policy. The stability of the exchange rate directly influences the overall macro-economy. Osorio et al. (2011) shows that economies with a relatively greater contribution from exchange rate and equity movements in the overall financial conditions, such as Hong Kong, Taiwan, and Singapore, tend to experience greater volatility in GDP growth. It may be thus not surprising that countries often follow practices of an active exchange rate management.1 Second, it is also important to note that exchange rate management and interventions occur mostly in emerging economies (Kriljenko and Iván, 2003). This phenomenon could be explained by a size effect: the intervention amounts have a relatively larger size to the level of market turnover in emerging economies compared to advanced economies (Levine, 1997). Furthermore, central banks in emerging economies have a greater information advantage over market participates. Owing to reporting requirements, central banks may infer aggregate order flow; they can actively use monetary regulation and operating practices (Scalia, 2008). Taiwan is an export-dependent economy with adopting a managed floating exchange rate system. Under this exchange rate regime, the NTD exchange rate is determined, in principle, by market supply and demand. However, when the market is disrupted by seasonal or irregular factors, Taiwan central bank will step in.2Yan and Shea (2005) confirm that the policy consideration, such as exchange rate stabilization, plays an important role in influencing the NTD/USD exchange rate trend, and has driven the Taiwan central bank to undertake significant intervention into the market. By assessing the size of the order imbalance effect on NTD/USD exchange rate return, the paper is aimed to provide central bank with guidance for the magnitude intervention in the future. Regarding the exchange rate dynamics issue, prior to the 1990s, analysis of the causes of exchange rate movements basically stems from macroeconomics arguments. However, plenty of empirical evidence shows that asset market models of exchange rate determination, using low frequency data on exchange rates and macroeconomic fundamentals, fail to explain exchange rate movements in the short run and can only indicate long-run trends.3 The failure of asset market approach in explaining exchange rate dynamics at shorter horizons and data availability on intraday foreign exchange transactions have led, in the last decade, to increasing attention to models of exchange rate determination based on market microstructure arguments. The principal empirical result of the new market microstructure approach to exchange rate determination is that order imbalance has the substantial explanatory power for exchange rate dynamics in the short term, from 5 min to daily interval, and even in the medium term, from weekly to monthly intervals.4 Order imbalance defined as the net of buyer-initiated and seller-initiated currency transactions (Lyons, 2001)5 is a measure of net buying pressure. The theoretical link between return dynamics and order imbalances mainly comes from two well-known channels of market microstructure theory.6 First, an information channel emerges when dealers adjust price in response to order flows that may reflect private information (e.g., Kyle, 1985, Glosten and Milgrom, 1985, Admati and Pfleiderer, 1988, Frömmel et al., 2008 and Danielsson et al., 2012; among others). Second, an inventory-control or liquidity channel emerges when dealers adjust price to control inventory risk due to order flows (e.g., Amihud and Mendelson, 1980, Ho and Stoll, 1983 and Brennan et al., 2012; among others). Both channels imply that buyer-initiated trades push price up, while seller-initiated trades push it down. Regarding the information-based channel in the field of foreign exchange rate, Breedon and Vitale (2010) distinguish two classes of traders: rational investors and unsophisticated customers. Rational investors represent all foreign exchange traders, such as dealers, hedge funds and of other actively traded funds, which have direct and full access to the two trading platforms: Reuters D2 and EBS (Electronic Broking System) systems. Unsophisticated customers correspond to traders, such as industrial corporations or institutional investors, which do not have direct access to these two trading platforms. These traders must phone up dealer brokers to get trading prices and complete a transaction. Thus, there exists asymmetric information between foreign exchange traders, so that, order imbalances can have the information content. Theoretical models that focus on the information contained in interdealer spot foreign trading activity are also found in Lyons (1995) and Perraudin and Vitale (1996).7 These models suggest that dealers receive private signals of future exchange rate dynamics from their customer order flow. The customer order flow may be informative for the following reasons. First, a dealer might have an intervening central bank as a customer and thus learn about future monetary policy and economic prospects from the central bank's order. Second, although information regarding future exchange rate fundamentals, such as interest rate and trade balance, is dispersed among individual customers, a dealer observes the aggregate trading behavior of a group of individual customers and thus may receive a signal regarding future exchange rate fundamentals. Thus, order flow can aggregate macroeconomic information for two reasons: (i) differential interpretation of news and (ii) heterogeneous expectations about future fundamentals. Several studies such as Berger et al. (2006), Dominguez and Panthaki (2006), Evans and Lyons, 2005 and Evans and Lyons, 2006, and Love and Payne (2008) investigate how order flow can reflect differential interpretations of news. These studies show that several macroeconomic indicators have statistically significant contemporaneous impact on order flow. Further, Rime et al. (2010) find that order flow's explanatory power stem partly from macroeconomic fundamentals. On the other hand, Rime et al. (2010) also focus on the role of order flow in capturing changes in heterogeneous expectations about future fundamentals. They directly investigate the transmission mechanism from real-time changes in expectations about future macroeconomic announcements to movements in exchange rates. As for the portfolio-balance channel, foreign exchange dealers are willing to absorb an excess demand (supply) of foreign currency from their customers only if compensated by a shift in the exchange rate (Evans and Lyons, 2002a and Breedon and Vitale, 2010). Flood (1994) and Lyons, 1995, Lyons, 1996 and Lyons, 2001 also suggest that dealers in the foreign exchange market tend to pass undesired positions along to another,8 thus giving rise to temporary misallocations of currency inventories which need to be compensated with a shift in expected returns. In the case of the large scale interventions, Evans and Lyons (2000) implicitly assume that every dollar of intervention translates into an equivalent dollar of order flow imbalance such that the inter-dealer market is left “holding” all the intervention trades. Nonetheless, it is possible that the order imbalance generated by intervention is rapidly translated into customer orders such that the inter-dealer market effectively passes the intervention on to customers and so dealers are left with little or no imbalance to trade amongst themselves. The actual impact of intervention on order flow is then used to calculate the predicted impact of order flow innovations on the spot rate. Evans and Lyons (2002a) are the first to find that exchange rate returns and signed order flows exhibit strongly positive correlated, and the explanatory power of signed order flows is dramatically large. That paper shows that the correlation between daily return and interdealer order flow is above 60% for deutsche mark/dollar and 40% for yen/dollar. Their results are found to be robust by subsequent studies on international currencies (e.g., Payne, 2003, Berger et al., 2006, Frömmel et al., 2008 and Breedon and Vitale, 2010). Therefore, what arouses our interest is whether our local currency, New Taiwan dollar/US dollar (NTD/USD), reaches conclusions similar to international currencies. Moreover, while a number of empirical evidence have documented the strongly positive return–order imbalance relation of exchange rate dynamics, there are few empirical results on the predictability of order imbalance with regard to exchange rates. It is traditionally known that exchange rates approximately follow a martingale so that future changes are unpredictable on the basis of publicly available information (Engle et al., 1990). Meese and Rogoff (1983) find that no available information is useful in predicting exchange rates out of sample better than a random walk model. Nevertheless, if order flow carries private information for exchange rate, then order flow should provide forecasting power for exchange rates. Recent studies, Evans and Lyons (2005), Gradojevic and Yang (2006), King et al. (2010), and Cerrato et al. (2011), show that customer order flows have predictive power for exchange rates. In contrast, studies of forecasting power of interbank order flows exist mixed conclusions.9Rime et al. (2010) find that interbank order flow does outperform the random walk in predicting daily exchange rates, while Segar and Taylar (2008) find that interbank order flow do not outperform the random walk when information on future fundamentals is unavailable. In this study, we try to investigate the predictability of interbank order imbalance for high frequency NTD/USD exchange rate, and use interbank order imbalance, one of public information, to test the efficient market hypothesis as defined by Fama (1970) and measure intraday speed of convergence to market efficiency. To sum up, we utilize the intraday dataset on NTD/USD exchange rate covering from 2 January 2008 through 31 December 2008 to explore the role of order imbalance in the high-frequency exchange rate dynamics. We study three empirical issues in sequence. First, we begin by analyzing the relation between return and order flows of NTD/USD foreign exchange under the time-series regression model and a time-varying GARCH(1,1) model by simultaneously incorporating order imbalance in the conditional mean and variance equations to capture NTD/USD dynamics. Second, we test the predictability of order imbalance on NTD/USD return by running time-series regressions of price movement on lagged imbalances as well as the out-of sample performance compared with random walk model. Third, to investigate the intraday efficiency of the interbank NTD/USD foreign exchange market, we trace the predictability of returns using order imbalances over various intervals. Our study relates to market microstructure argument of exchange rate determination and makes marginal contributions to the existing intraday NTD/USD dynamics studies as follows. First of all, we investigate the explanatory and predictability of order imbalances using the dataset that covers recent trading records. Previous studies are limited to the trading records before 2001.10 Our new dataset will be helpful for generating more reliable results on the intraday NTD/USD dynamics following the further liberalizing in the local foreign exchange market. Secondly, the unavailability of the order flow led prior studies to concentrate on the price–volume patterns and determinants of intraday volatility, while we argue that the direct relationship between order imbalances and returns should consider linkages with volatility. Third, the findings on the significant impact of order flow on the exchange rate allowed the central bank to effectively facilitate exchange rate stability using order intervention, as central bank has a relatively larger size in interbank market. Furthermore, through monitoring the order imbalance information (both fundamental and non-fundamental information), the central bank could potentially use moral suasion or other means to discipline any private banks that might be viewed as contributing to a speculative attack. The main results of the study are stated as follows. First, contemporaneous order imbalances have positive and significant influences on high frequency NTD/USD returns. Second, regarding the predictability of order imbalances for currency returns, there exists a significantly negative relation between lagged-one order imbalance and current NTD/USD returns on the intraday analysis. The negative relation could be explained by two possible reasons: (i) the price stabilization mechanism executed by the Taiwan central bank: the intervention against the wind might lead to NTD/USD price reversals following the lagged imbalances. (ii) the existence of overreaction phenomenon in the interbank market. Further, by the out-of-sample test, we find that the performance of using order imbalance in forecasting exchange rate movements is better than that of using random walk in daily and intraday data. Third, for the study on intraday speed of convergence to market efficiency, we find that the significance in lagged order imbalances on predicting returns are up to 60 min during NTD depreciation period while up to 15 min during NTD appreciation period. The remainder of our study is organized as follows. Section 2 describes data. Section 3 presents the empirical methodology, including the time-series regression model and the GARCH(1,1) model to test the return–order imbalance relation, the predictability of order imbalance and the market efficiency. Section 4 discusses the empirical results and analyses on high frequency NTD/USD dynamics. Section 5 concludes.

In this study, we utilize the intraday dataset on NTD/USD exchange rate covering from 2 January 2008 through 31 December 2008 to explore the role of order imbalance in the high frequency exchange rate dynamics. We study three issues in sequence. First, we begin by analyzing the relation between return and order flows of NTD/USD foreign exchange under the time-series regression model. We find that contemporaneous order imbalances have positive and significant influences on high frequency NTD/USD returns. To discuss whether the return–order imbalance relationship arises due to volatility increases, we propose a time-varying GARCH(1,1) model by simultaneously incorporating order imbalance in the conditional mean and variance equations. The size of current order imbalance coefficient, which measures the price impact of order flow, in GARCH model is generally smaller than that in the linear time-series regression. This tells us some of the explanatory power in the OLS regression attributes to volatility. Second, we provide the evidence of the predictability of order imbalance on NTD/USD return by running time-series regressions of price movement on lagged imbalances. When contemporaneous imbalances are not included in the regression, lagged-one order imbalances have significantly negative impacts on current returns. We provide two possible explanations for the negative relation. First, the price stabilization mechanism executed by the Taiwan central bank: the news reported about the central bank behaviors in our sample period seems to support this view. Second, the existence of overreaction phenomenon in the interbank NTD/USD market: We find that intraday price reversals following large price changes at the market open in our sample, which are consistent with the evidence from other financial markets (e.g. Atkins and Dyl, 1990, Fung et al., 2000 and Grant et al., 2005). Furthermore, by the out-of-sample test, we find that using the interbank order flow in the time-series regression model does outperform the random walk model both in the daily and intraday data. However, comparing the out-of-sample accuracy of time-series regression models without volatility and GARCH model with volatility in forecasting the NTD/USD returns, the performance of GARCH model is statistically indifferent to that of OLS models. This means that volatility affects explanation power of order imbalances on returns, while it does not improve the predictability for our sample. Third, to investigate the intraday efficiency of the interbank NTD/USD foreign exchange market, we trace the predictability of returns using order imbalances over various intervals. Overall, our results show that the weak-form efficiency appears to prevail over intervals from 15 to 60 min in the interbank NTD/USD exchange rate market. This rapid convergence to efficiency could be explained by the interactions of three distinct groups in the interbank exchange rate market. Order imbalances firstly arise from end-users with various demands. Order imbalances are positively autocorrelated, which is related to splitting orders (Kyle, 1985), and/or herding effect (Hirshleifer et al., 1994). Then dealers trade each other and react to initial order imbalances by altering quotes away from fundamental value to unwind the undesired positions. Finally, contrarian dealers (Wan and Kao, 2009), dealers with information learned from previous orders (Romeu, 2003) conduct countervailing trades in the direction opposite to the initial order imbalances. In summary, our empirical findings on the significant impact of order flow on the exchange rate and the size of the order imbalance effect on exchange rate return could allow central bank to effectively facilitate exchange rate stability using order intervention, as central bank has a relatively larger size in interbank market and could monitor the order imbalance information via reporting requirements.