Patterns of financial crisis: AUD/JPY in 2007-2008.

User Rating: / 0
PoorBest 
Written by Forex Automaton   
Wednesday, 28 January 2009 15:14

In this series of posts I look for specific changes in the autocorrelation structure of leading forex exchange rates, related to the global financial crisis. The Holy Grail of automated trading system development is to find "stable imperfections" -- deviations from market efficiency which remain ever-present, or present at predictable times, or present most of the time without turning into their opposites the rest of time, or at lest present long enough for the AI to detect and "learn" them, use them, and "unlearn" them if they are no longer there without disastrous losses. From this point of view, you can share our interest in the "crises" of all sorts -- these are the periods when some of the rules that worked in the past no longer do. A burning question is whether the episodes of panic are characterized by more order or less order, compared to "normal" times, and whether this is the "same" order or the rules of order are changed. The previous case studies, dealing with EUR/JPY, EUR/USD, GBP/USD, had a common conclusion: the rules of crisis are different, and seem to advance a particular correlation pattern, which I called "bipolar disorder". The case of GBP/JPY was not so clear, possibly because GBP/JPY was seen to be possessed by "bipolar disorder" even in a late 2002 -- early 2008 analysis. In the case of AUD/JPY, the "bipolar disorder" feature was seen to be most pronounced in 2004 (cheap money period) and to recede already in 2007, which made me associate it tentatively with carry trade. For these high interest differential exchange rates, the "bipolar disorder" seems to be the hallmark of "good times", times of high risk tolerance. This complicates the interpretation. To make things even more complex, for the Aussie (as for the Sterling) the crisis brought about a considerable reduction of the interest rates, thus changing the playground as far as the rate differential games are concerned. Nevertheless, below are the details for AUD/JPY.

Evolution of AUD/JPY exchange rate during the financial crisis, hour.

Fig.1:AUD/JPY during the financial crisis, hour time scale. Time axis is labeled in MM-YY format and spans the interval from August 2, 2007 through December 31, 2008.

As before, the time interval starting in August 2007 is split into two parts, the "moderate" and the high volatility one, whose features are then compared. I define the visible phase of the present financial crisis to begin on August 16, 2007, the day of Countrywide Financial near-bankruptcy event, followed by an extraordinary half-percent Fed discount rate cut next day. This study covers 74 weeks from August 2, 2007 through December 31, 2008. The sub-range of extreme volatility (as will be seen in Fig.4) can be roughly defined as the last 18 weeks of this 74-week range. Thus, I divide the time interval under study in two unequal parts, those of (relatively speaking) low and high volatility. The low volatility phase is from August 2, 2007 through August 27, 2008. The high volatility phase is from August 28 through the end of 2008. In this study, I only look at trading activity taking place from 1am to 1pm New York time, since the experience shows it to be the richest in non-trivial correlations.

Autocorrelation of logarithmic returns in AUD/JPY,  European trading hours, hour scale, from August 2, 2007 to August 28, 2008. Autocorrelation of logarithmic returns in AUD/JPY,  European trading hours, hour scale, from August 28, 2008 through December 31, 2008.

Fig.2: Autocorrelation of logarithmic returns in AUD/JPY for the European (Eurasian) trading shown against the backdrop of statistical noise (red). Top panel: the measurement time range is for the relatively low volatility phase of the crisis, from August 2, 2007 through August 27, 2008. Bottom panel: same for the high volatility phase, from August 28, 2008 through the end of 2008. The noise is obtained from martingale simulations based on the recorded volatilities of AUD/JPY in the trading hours under study for the period. The noise is presented as mean plus-minus 1 RMS, where RMS characterizes the distribution of the correlation value obtained for each particular bin by analyzing 20 independent simulated uncorrelated time series of the same average volatility.

Evolution of AUD/JPY autocorrelation peak structure during the financial crisis, hour.

Fig.3: Evolution of AUD/JPY autocorrelation peak structure during the financial crisis, hour time scale. Time bin is two weeks wide. The peak structure is represented by three correlation values: the one for the zero lag (essentially a volatility measure) downscaled by 10 for easier visual comparison, the one at one hour lag (just discussed) and the one at two hour lag. Time axis is labeled in MM-YY format and spans the interval from August 2, 2007 through December 31, 2008. Only trading hours from 1am to 1pm New York time (European trading hours), usually rich in non-trivial correlations, are included.

In Fig.2, the pattern in the vicinity of the zero time lag bin looks consistent in both top and bottom panels. Fig.3 presents the evolution of the correlation structure in the vicinity of zero time lag, representing the correlation structure as a triplet of correlation values: those at zero, one and two hour lags. The increased volatility shows up as the increase in the magnitude of all these values, with variance (a measure of volatility) being fairly well represented by the magnitude of the zero time lag value. The time period of dramatically higher volatility covers the last 9 bins in Fig.3, which contain 18 weeks.

The magnitude of the negative one-hour lagged correlation (hatched in red in Fig.3) is seen to be about factor of 10 below the variance (variance downscaled by 10 is hatched in gray in Fig.3). Out of the last 9 time bins in Fig.3, 7 have negative signal at one hour lag, usually associated with "bipolar disorder". Despite that, the one hour signal in Fig.2, bottom panel, is only barely above the noise -- in part due to the contribution from the last two weeks of October when the positive correlation was so wide that the one hour lag bin had a positive signal. Interestingly, the moment when AUD/JPY hit the long-time bottom at around 56 and turned around belongs to that time bin. GBP/JPY did something very similar, with a very similar correlation pattern, even though for GBP/JPY the bottom turned out to be more local.

The data used are from the period 2002-08-02 00:00:00 to 2009-01-01 00:00:00, New York time.

Bookmark with:

Deli.cio.us    Digg    reddit    Facebook    StumbleUpon    Newsvine
Last Updated ( Monday, 04 January 2010 12:41 )