تناوب روزانه، تقویم و اعلام اثرات نوسانات نرخ ارز در یورو

This paper provides an analysis of intraday volatility using 5-min returns for Euro–Dollar, Euro–Sterling and Euro–Yen exchange rates, and therefore a new market setting. This includes a comparison of the performance of the Fourier flexible form (FFF) intraday volatility filter with an alternative cubic spline approach in the modelling of high frequency exchange rate volatility. Analysis of various potential calendar effects and seasonal chronological changes reveals that although such effects cause deviations from the average intraday volatility pattern, these intraday timing effects are in many cases only marginally statistically significant and are insignificant in economic terms. Results for the cubic spline approach imply that significant macroeconomic announcement effects are larger and far more quickly absorbed into exchange rates than is suggested by the FFF model, and underscores the advantage of the cubic spline in permitting the periodicity in intraday volatility to be more closely identified. Further analysis of macroeconomic announcement effects on volatility by country of origin (including the US, Eurozone, UK, Germany, France and Japan) reveals that the predominant reactions occur in response to US macroeconomic news, but that Eurozone, German and UK announcements also cause significant volatility reactions. Furthermore, Eurozone announcements are found to impact significantly upon volatility in the pre-announcement period.

The analysis of high frequency financial market data in recent years has revealed a distinctive intraday volatility pattern, or ‘intraday periodicity’.1 Attempts to prove that this periodicity in volatility is neither sample nor market specific have led to investigation of a wide variety of alternative asset markets, and this literature has reinforced the importance, robustness and regularity of intraday volatility patterns across global markets and financial instruments.2 The foreign exchange market has attracted particular attention in this context because the separation of the foreign exchange market across regional financial centres and disparate time zones permits continuous trading and offers an interesting and challenging context for the modelling of intraday periodicity. That is, the microstructure of the foreign exchange market dictates a 24 h pattern to intraday volatility governed by trading activity in the world's major financial centres, whereby volatility increases at market openings and when trading in the most active centres overlap, whilst this inherent pattern is disrupted by severe spikes immediately following the release of macroeconomic news.3 The filtration of high frequency returns volatility through modelling of the underlying pattern is therefore an essential precursor to the modelling of volatility, and any empirical analysis of, for example, the impact effects and dynamic responses associated with news. Andersen and Bollerslev (1998) provide a robust econometric methodology for capturing the distinct volatility components and isolating macroeconomic announcement effects. This involves adopting a deterministic intraday volatility pattern to capture high frequency volatility periodicity, and imposing a predetermined volatility response pattern associated with calendar and other effects. The filtration of absolute returns by such an intraday periodicity component, estimated by a Fourier flexible form (FFF), and standardisation by an estimated daily GARCH component to account for persistence at lower frequencies, reveals interesting patterns in the correlogram of absolute returns that are invisible prior to the periodic filtering. As well as successive U-shaped intraday patterns, autocorrelations at the daily frequency show a cyclical pattern, and decay slowly over the first four days only to increase slightly at the weekly frequency, signalling a minor day-of-the-week effect. The combination of recurring cycles at the daily frequency and a slow decay in the autocorrelations are explained by the joint presence of the pronounced intraday periodicity and strongly persistent daily conditional heteroscedasticity. Whilst the FFF method has also been applied by Andersen et al. (2000) and Bollerslev et al. (2000) to different market settings, very few other studies tackle fully the complexity involved in the modelling of intraday volatility of exchange rates, electing instead to filter intraday periodicity by standardising the chosen measure of volatility using the average absolute or squared return for a particular intraday interval over the sample period. Such techniques do not lend themselves to the further modelling of macroeconomic announcement effects, which forms a key element of the empirical investigation here, since much of the announcement effect in the intraday interval following an announcement is removed by such an arbitrary filtering approach. Whilst the FFF method is parsimonious and allows for smooth volatility dynamics, it is rigid in functional form, and imposes a smooth cyclical pattern in the characterisation of intraday periodicity. An interesting alternative to the FFF method is provided by the cubic spline method previously utilised by Engle and Russell (1998), Zhang et al. (2001), Taylor, 2004a and Taylor, 2004b and Giot (2005) in the context of autoregressive conditional duration models applied to irregularly spaced transaction data, but which has yet to be applied to foreign exchange data. This alternative method allows different cubic spline functions to be estimated between selected points (termed ‘knots’) in the periodic cycle, such as the various market opening and closing times in 24 h foreign exchange trading, and offers the potential to more closely match the fitted intraday periodic pattern with the known times of opening and closing in the principal markets, so potentially enhancing the efficiency of tests for calendar and macroeconomic announcement effects. This study contributes to the existing literature in four main ways. First, it considers the intraday periodicity in the exchange rates of the Euro against the US Dollar, Pound Sterling and Japanese Yen, which constitutes a new market that has yet to be investigated in this econometric framework. The modelling of intraday patterns in the volatility of Euro exchange rates has, to the best of our knowledge, not yet received full empirical attention in the literature, and the robustness of previous results for the pattern of intraday volatility and the impact of various calendar effects on volatility, which have previously typically been conducted for US dollar exchange rates, therefore remains to be addressed.4 Second, the study compares the two alternative techniques for capturing the intraday volatility pattern described above, namely the FFF method and the alternative cubic spline specification. Third, the empirical methodology adopted allows for an assessment of the statistical and economic significance of various time and calendar effects associated with the opening of related markets, winter and summer periods, public holidays, and potential day-of-the-week effects. Fourth, we consider the empirical significance of macroeconomic announcements grouped by country of origin emanating from the US, Eurozone, Germany, France, UK and Japan on exchange rate volatility during the intervals prior to, immediately following, and after those announcements. This constitutes a wider range of macroeconomic announcement sources than considered hitherto in the literature. The remainder of the paper is structured as follows. Section 2 describes the data and the results of some preliminary analysis, including log-periodogram estimates of the degree of fractional integration in the data. Section 3 describes the econometric modelling approach and Section 4 explains the two intraday periodicity filters applied. Section 5 reports the results of applying these filters to the exchange rate data and discusses the relative success of the two procedures in replicating the average periodicity. Section 5 also reports and discusses the statistical and economic significance of various calendar and macroeconomic announcement effects. Section 6 summarises and concludes the paper.

The components of high frequency returns volatility are not only significant and interesting in statistical and economic terms, but the identification and accurate modelling of their dynamics is also a crucial precursor to their formal econometric modelling using models of the GARCH or stochastic volatility varieties, and for the analysis of macroeconomic announcement effects on foreign exchange market volatility. Using a 19-month sample of 5-min returns for Euro–Dollar, Euro–Sterling and Euro–Yen exchange rates, and therefore a new market setting, this study confirms a 24 h pattern for intraday volatility, with volatility rising at the opening and overlapping of trading activity in the world's major financial centres. Whilst previous studies of this type commonly filter intraday volatility by fitting a Fourier flexible form (FFF) to the intraday pattern, this study compares the performance of the FFF with an alternative cubic spline approach. A particular advantage of the cubic spline method over the FFF approach is that it does not impose a smooth pattern on intraday volatility, but allows sharp peaks and troughs. Consequently, the cubic spline method is able to provide a closer characterisation of the changing nature of periodicity through the 24 h foreign exchange trading cycle, and the sharp changes in volatility associated with market opening and closing times. A clear example is provided by the peak in volatility during morning trading in Europe and the UK. The superiority of the cubic spline approach in this regard is also confirmed by, for example, the absence of significant timing effects associated with the opening of trading exchanges in Tokyo under the cubic spline method but not under the FFF approach. Further analysis of various potential calendar effects associated with seasonal chronological changes reveals that these calendar effects present interesting deviations from the average intraday volatility pattern, with notable volatility increases associated with the change to daylight saving time, Tuesdays and Thursdays. However, such intraday timing effects are in many cases only marginally statistically significant and are insignificant in economic terms, as judged by their effect on volatility over the entire horizon of the response and against cumulative absolute returns measured over a typical day. Concerning the robustness of our findings to the two alternative intraday periodicity filters employed, the cubic spline model results imply that announcement effects are larger immediately following the announcement and are far more quickly absorbed into exchange rates in the post-announcement period than is suggested by the FFF model results, which underscores the advantage of the flexibility of the cubic spline approach in permitting intraday volatility periodicity to be more closely identified. Finally, examination of the exchange rate volatility response to macroeconomic announcements emanating from the US, Eurozone, UK, France, Germany and Japan reveals that the predominant reaction of volatility, across all three exchange rates, occurs in response to US macroeconomic news, but also reveals that, on average, Eurozone, German and UK announcements also cause significant reaction in exchange rate volatility. Whilst the majority of such reactions in volatility occur after the announcement, Eurozone announcements are found to impact significantly upon volatility in the pre-announcement period. Possible explanations for this finding, which might usefully be addressed in future research, include information leakage in the period leading up to the announcement, departures from scheduled release times, or positioning by traders in the run-up to ECB interest rate announcements.