وابستگی متقابل و پویایی در بازار آتی پول : تجزیه و تحلیل چند متغیره داده های معاملات طول روز
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|14343||2001||26 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Banking & Finance, Volume 25, Issue 6, June 2001, Pages 1161–1186
This paper investigates long-term interdependencies and short-term dynamics in currency futures utilizing intraday data for six major foreign currencies: the British Pound, Deutsche Mark, Swiss Franc, Australian Dollar, Canadian Dollar, and Japanese Yen. Lack of cointegration (CI) among the foreign exchange futures is found to be the prevailing mode of behavior, but some temporary deviations from the no-CI condition are detected. There is a notable overlap between detected CI relationships and the timing of policy changes, world events, and regime shifts, indicating that the observed CIs are event-driven. The robustness of the CI results is checked with respect to variations in the model, lag structure, data period, sample horizon, and currency basket grouping. Impulse–response functions (IRFs) reveal that currency markets are in general efficient and absorb new information within the day. The interdependence among currencies is found to be asymmetric.
Investigation of long-term interdependencies and short-term price dynamics across and within currency markets provides valuable insight for traders and policy makers since these patterns signal potential informational inefficiency and hence, possible asset mispricing. Traders are interested in interdependencies and price dynamics because they may offer profitable arbitrage opportunities, while policy makers are eager to understand the nature of the information transmission process in order to coordinate policy internationally, and to design remedies for market imperfection problems such as delays between information arrival and valuation process. King and Wadhwani (1990) provide a theoretical foundation for interdependence among financial markets. According to the model proposed by these authors, attempts by the rational economic agents in each market, to infer from price changes in other markets, creates a contagion across the markets. In this framework, price changes in any one market will be sensitive to price changes elsewhere, as well as the market fundamentals, and mistakes and idiosyncratic changes in any one market will carry over to the rest. The King and Wadhwani view is consistent with the “Meteor Shower” hypothesis put forward and empirically validated by Engle et al. (1990), according to which volatility spills over across markets, rather than remain market-specific in character. In addition, King and Wadhwani present empirical evidence that as volatility increases contagion strengthens. Harvey and Huang (1991) maintain that although disclosure of private information through trading can partially explain the fluctuations in the currency futures markets, as proposed by King and Wadhwani, it is the public announcement of macroeconomic news that constitutes the prime motivator of the movements in the currency futures markets and the most likely factor to induce volatility in these markets. In the recent years, the long-term interdependence among financial markets has been widely investigated using the cointegration (CI) technique. It is well known that interdependencies described by a CI relationship across prices can significantly improve the prediction of future price values. Granger, 1981 and Granger, 1986 provides a link between CI and efficiency. According to Granger, in informationally efficient markets, prices of assets should not be cointegrated; if asset prices are cointegrated and market participants can profitably use this information, markets should be considered inefficient. Granger’s proposition opened a new venue for testing market efficiency with CI tests serving as the standard tool. Subsequently, numerous studies examined presence of CI across currencies, using daily or monthly data, and pair-wise or system-based tests. The findings of these studies are mixed. For example, MacDonald and Taylor (1989) and Hakkio and Rush (1989) apply the Engle–Granger pair-wise procedure to a number of currencies (vis-à-vis the British Pound) and fail to reject the absence of pair-wise CI. Baillie and Bollerslev, 1989 and Sephton and Larsen, 1991, and Copeland (1991) employ the Johansen (1988) procedure to conduct similar tests: while the former authors find evidence in favor of pair-wise CI in currency markets, Copeland fails to reject the null hypothesis of no-CI. 1 More recently, Diebold et al., 1994 and Lajaunie and Naka, 1992, and Lajaunie et al. (1996) have utilized the more general system-based Johansen (1991) procedure in order to test the absence of CI in a basket, rather than a pair, of currencies. The findings of these authors also lend support to the absence of CI among the currencies considered and the prevalence of efficiency. The link between CI and efficiency, however, has been the subject of dispute. In particular, Baillie and Bollerslev, 1989 and Baillie and Bollerslev, 1990 and Crowder, 1994 and Crowder, 1996 show that the correspondence between CI and efficiency is weak and that empirical finding of CI may indeed stem from sources other than inefficiency. Specifically, Crowder (1996) identifies four sources for the finding of CI: (a) markets are inefficient and traders are indeed wasting valuable information, (b) markets are efficient but there exist some omitted factors, such as a risk premia, or regime switches, that manifest themselves as CI, (c) markets are inefficient but agents are ignoring the information from the error correction models because it cannot engender significant profits, and (d) finding of CI is due to questionable statistical properties of the tests. According to this scenario, even if the implemented CI tests are powerful and the results do offer profitable improvements in prediction, the finding of CI can present a challenge to market efficiency only if the sample period is free of significant world events, regime shifts, and policy changes. Thus, it is essential that, when conducting CI tests, one identifies the political and economic events in the sample period and determines whether the timing of the observed CI relationships corresponds to these events. Most studies fail to account for all the aforesaid sources of CI, and in particular, the event-driven nature of the CI results. A contribution of this study is that it examines the link between world events and observed CI relationships. 2 Another feature of the existing studies of currency markets is that they employ relatively low frequency and long-horizon data when investigating CI relationships, presumably on the grounds that CI is a long-term equilibrium phenomenon.3 However, a no-CI result based on daily data does not rule out the presence of information inefficiencies within the day. This issue merits attention because it is widely known that absorption of new information in stock, currency, and futures markets occurs in horizons shorter than a day, often even less than an hour. For example, Jeong (1994) points out that “the speed of information dissemination in global stock markets is literally instantaneous, except for the time necessary to interpret the information”. Along similar lines, Becker et al. (1992) find that the Japanese stock market reacts within the hour to information shocks originating from the US. 4 Ederington and Lee (1993) and Tanner (1997) corroborate these findings. Ederington and Lee (1993) examine the clearing process in the interest rate and currency futures markets in response to news. Based on their empirical results, they argue that traders with immediate access to these markets form expectations of the new prices almost immediately after the news occurs, and the actual prices adjust to this expectation within one minute. Once news details become available, new traders enter the market, and all traders reassess the implications of the news. As a consequence, price fluctuation will persist at a considerably higher level than normal for a window of 15 minutes and will continue to be slightly higher for several hours. The final equilibrium price, however, will not deviate substantially from the one reached within the first minute. The findings of Tanner (1997) are in general consistent with the Ederington and Lee’s scenario but denote a slower speed of convergence. According to Tanner, the reaction of the currency markets to the news about trade deficits takes less than 30 minutes to complete, while the reaction to the inflation news requires 3.5 hours to be fully absorbed. Even in the latter case, however, the bulk of the effect does occur within the first hour of the news announcement. The accounting literature provides some contrasting evidence to the findings of Ederington and Lee (1993) and Tanner (1997). For example, Bernard et al. (1989) and Bernard and Thomas (1990) find that US stock returns respond to the earning announcements with a delay of up to three days. Similarly, Hew (1996) reports supporting evidence for the presence of analogous delays for the UK small firms, though not as much for the large firms.5 Based on the extant literature, while it is agreed that CI tests pertain to the “long-run”, as opposed to the “short-run”, analysis, the adequate time horizon needed for the establishment of a long-run equilibrium in currency futures markets is understood to be relatively “short”.6 Under this condition, the use of high-frequency data over short horizons can be said to make tests regarding currency interdependence more incisive. Two arguments support this position. First, for markets with high liquidity and rapid adjustment such as currency futures, inefficiencies can be eliminated and new equilibria can be achieved within the intraday trading periods. Hence, the use of daily data may not allow inefficiencies to be manifested. This problem can be solved by the use of high-frequency data, which allows equilibria to be demarcated over shorter time spans. The shorter horizon also permits a longer sequence of equilibria to be incorporated into the analysis, for a given length of time. Second, with shorter sample horizons, the effect of regime shifts, government intervention in the market, and other omitted factors can be isolated and restricted to the event period, rather than being allowed to cloud the overall picture (Crowder, 1996). In addition to long-term CI relationships, the short-term dynamics of information transmission across markets also deserves attention due to its implications for informational efficiency and international risk diversification. Informational efficiency implies that the time interval between new information arrival and price adjustment should be relatively short. Hence, to examine informational efficiency, one can resort to the impulse–response function (IRF) analysis as a complement to CI. IRFs measure the effect of a shock in a given currency on other currencies over time and thereby they determine how fast the newly arrived information gets incorporated into the exchange rates. If markets are efficient the IRFs will “die out” quickly, displaying the lack of an intermarket effect over time. The degree of interdependence among the currencies can be used as a rough indicator of integration among the respective markets. If the currency markets in the system are integrated, then they will show a high degree of interaction. Otherwise, at least some of the currencies will be isolated from the rest and will be determined predominantly by their own internal dynamics. From a risk management view point, information spillover across currencies deserves attention because when spillover is substantial the motivation for holding a diversified foreign exchange portfolio will be curtailed. This effect becomes even stronger, when the speed of the international shock transmission increases, further shortening the time interval over which international diversification benefits can be secured. In addition, spillover asymmetry among currencies may have significant implications on portfolio management decisions. For example, if currency A is unaffected by currency B but B is predominantly determined by A, diversification benefits would be limited for portfolio managers investing in A, while not to investors in B. This is so because under these conditions the currency shocks in B will have no repercussions on A. Much work has been done in regard to the dynamics across markets using the IRF technique. For example, Eun and Shim (1989) and Karolyi (1995) have shown that information spillover effects do exist between stock market returns, may be asymmetric, and tend to be transmitted mostly fromlarger markets (such as the US) to others. Application of the IRF methodology to currency markets, however, has been more limited. Since IRFs provide complementary information concerning market dynamics, a combination of the IRF and CI results produces a more complete picture of the presence of spillover in currency markets. This study aims to remedy some of the shortcomings of the extant literature about delineating the long-term equilibrium relationships and the short-run dynamics in currency futures markets. The question of intertemporal stability of the CI results raised by Sephton and Larsen (1991), and the event-driven nature of CI results given prominence by Crowder (1996) are examined in the framework of the system-based Johansen (1991) procedure. CI test results are complemented with the IRF analysis in order to to gain additional insight on the existence, speed, and symmetry of information spillovers. Given the shortcomings of the CI tests, when used in isolation, in assessing informational efficiency of markets, the combined evidence based on CI and IRF analyses presented here yields a more persuasive account of market dynamics and information flow in currency markets. The rest of the paper is organized as follows: Section 2 discusses the data. Sections 3 and 4 introduce the methodology and present the CI results, respectively. Section 5 conducts sensitivity analyses with respect to the choice of model, lag structure, data frequency, and breadth of the currency basket. Section 6 discusses the findings based on the IRF analysis, and Section 7 summarizes and concludes the paper.