USD/CAD intraday seasonality overview, 2003-2010

User Rating: / 2
Written by Forex Automaton   
Thursday, 30 December 2010 15:40
Article Index
USD/CAD intraday seasonality overview, 2003-2010
Variances and correlations
All Pages

The regime change from mean reversion to trend following is seen to occur daily in USD/CAD. Like in other pairs studied so far, the statistically preferred time to bet on a trend reversal in USD/CAD is the morning hours of the Asia-Pacific trading session. Evidence for that comes from time dependence of autocorrelations of logarithmic returns with 1-hour lag. The hour ending at 17:00 CET (11am ET) favors trend following.

As always, the quantities we are going to look at are not the actual low, high and close. Since prices are always positive, they are trivially correlated; this feature is absent in the correlations of the so-called logarithmic returns (or logarithmic increments) which are the ratios of price levels (low, high, close) to the values they had during the previous hour-long time interval.

For years 2003 through 2009, the entire yearly data sets are used. For 2010, the data from the beginning of the year to November 1 are used.

Central European time is chosen because it allows one to split the forex week into five 24-hour long non-interrupted trading days.

To look for seasonality effects in first-order statistics (averages), we average logarithmic returns for each hour of the day separately using profile histograms. The resulting histograms are plotted in Fig.1.

USD/CAD averaged log return vs CET hour of the day, in 2003, 2004, 2005, 2006 1.1 USD/CAD averaged log return vs CET hour of the day, in 2007, 2008, 2009, 2010 1.2 USD/CAD averaged log return vs CET hour of the day, in 2003-2010. All years aggregated. 1.3 USD/CAD averaged log return vs CET hour of the day, in 2003-2010. All years aggregated. Coherent time window. 1.4

Fig.1. Averaged logarithmic return vs hour of the day in CE time (see time conversion table to other time zones) for USD/CAD. 1.1: Years 2003, 2004, 2005, 2006. 1.2: Years 2007, 2008, 2009, 2010. 1.3: All years added. Vertical bars show estimates for the precision of the mean. 1.4: Years 2003-2010, showing the time window of the highly consistent behavior only. Hatched bands in the plots are twice as wide as the measurement's precision of the mean. The mean is the middle of the band. Because hourly returns do not have Gaussian distributions, the usual quantitative rules of thumb about confidence intervals and confidence probabilities do not apply here. Nevertheless, stability of the effect in panel 1.4 is hard to deny.

Fig.1 presents hourly "seasonal" averages of the hourly logarithmic return in USD/CAD. There is a considerable variation among the patterns, and like EUR/USD and GBP/USD, the 00:00--4:00 slot (ends included, hour quoted indicates the end of the time interval for which the returns are averaged) seems to be the time window where the same pattern of non-zero residual average tends to reproduce itself year after year. Fig.1.3 reveals a non-zero residual pattern surviving after aggregating the yearly data.

Last Updated ( Thursday, 10 February 2011 09:46 )