Wednesday, February 20, 2013

Losing control?

This is the monthly price of gold for the last ~12 years (145 months), in log scale.  X axis is in months since ~Jan 2001:

This is the % change of a given month over the prevous month:

There seem to be more outliers in recent months compared to earlier months.  Is this true?  The following graph shows the standard deviation of MoM %changes from above, calculated over a 20-month sliding window (the window size is arbitrary):

The huge peak around month 80-90 is driven by the outlier down month, Sept '08, circled in red in the 2nd plot above.  If we remove that data point, here is the resulting trend of standard deviations over 20-month sliding windows:

Still increasing pretty steadily, wouldn't you say?  Keep in mind this is standard deviation of percent month/month change, so it's already normalized for price.  Volatility is clearly increasing.