Patterns of financial crisis: USD/CAD in 2007-2009.

User Rating: / 1
PoorBest 
Written by Forex Automaton   
Wednesday, 08 April 2009 16:47

With USD/CAD, I am continuing the series of reports focusing in the time evolution of the forex correlation shapes during the present financial crisis. Extending the time coverage up to the end of March, I see the need to make the picture a bit more complex with three, rather than two phases with considerably different volatility level, since the volatility is seen to abate at the end of January 2009. The bipolar correlation pattern, a major subject of this research, is seen to disappear during the peak of volatility, but reappear later.

I define the visible phase of the present financial crisis to begin on August 16, 2007, the day of Countrywide Financial near-bancruptcy event, followed by an extraoridinary half-percent Fed discount rate cut next day. This study covers 88 weeks from August 2, 2007 through March 26, 2008. The sub-range of extreme volatility, Phase 2, is defined, according to the volaitlity history in Fig.2, to begin on August 28, 2008 and last till January 29th, 2009 -- this choice of dates is somewhat arbitrary and is motivated in part by desire to have two-week long bins or integration intervals. 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.

USD/CAD chart 2007-2009, hour

Fig.1:USD/CAD during the financial crisis so far, hour time scale. Time axis is labeled in MM-YY format and spans the interval from August 2, 2007 through March 26, 2009.

Evolution of USD/CAD autocorrelation peak structure during the financial crisis, hour.

Fig.2: Evolution of USD/CAD 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 March 26, 2009. Only trading hours from 1am to 1pm New York time (European trading hours), usually rich in non-trivial correlations, are included.

 

Fig.2 shows the corresponding evolution of the autocorrelation of logarithmic returns in the vicinity of zero time lag. The correlation structure is represented 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.

Most of the ForexAutomaton articles on the "patterns of financial crisis" turned out to be mainly about the bipolar pattern. By this I denote a tendency to form quickly alternating rises and falls on next-hour time scale, more pronounced than in a fully unpredictable time series of the same volatility. The pattern shows up as negative deeps surrounding the zero-time lag peak, and the way to quantify it in the hour scale analysis is to look at the correlation value at one hour lag -- presented in red in Fig.2. Looking at Fig.2, the bipolar pattern is seen wherever the red-shaded content is confidently in the negative area. It's remarkable how stable this pattern is initially during Phase 1 and during Phase 3. The effect is seen to disapper during the high volatility phase, the Phase 2. This conclusion is best illustrated by presenting time-integrated autocorrelations for the three time intervals under study.

Autocorrelation of logarithmic returns in USD/CAD,  European trading hours, hour scale, from August 2, 2007 through August 27, 2008. Autocorrelation of logarithmic returns in USD/CAD,  European trading hours, hour scale, from August 28, 2008 up to  January 28, 2009. Autocorrelation of logarithmic returns in USD/CAD,  European trading hours, hour scale, from January 29, 2009 up to  March 26, 2009.

Fig.3: Autocorrelation of logarithmic returns in USD/CAD 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 up to January 29, 2009. The noise is obtained from martingale simulations based on the recorded volatilities of USD/CAD 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.

Fig.3 shows that the bipolarity is gone or reduced during the high volarility phase. Fig.4.1 of the Annual Progress Report demonstrates how uncommon such a behaviour is in forex. The closest analog is perhaps EUR/CHF and AUD/JPY where the high volatility was found to merely preserve the shape of the correlation, as opposed to enhancing the bipolarity.

The data used are from the period 2002-08-02 00:00:00 to 2009-03-26 00:00:00, New York time.

Bookmark with:

Deli.cio.us    Digg    reddit    Facebook    StumbleUpon    Newsvine
Last Updated ( Monday, 04 January 2010 12:37 )