Thursday, September 9, 2010

VIX Seasonalities & Return Distribution Patterns

Ulises Carcamo Carcamo's PhD Thesis at University of Canterbury (Sept. 4th earthquake) points out a few interesting phenomenons about the VIX and OEX returns from 1988 to 2002. The math applied is relatively fundamental, as this is not an applied math doctorate, a basic understanding of statistics would be sufficient to read this paper.

1. The VIX (CBOE Volatility Index) likely holds some seasonality over time, verified via several statistical means. VIX tests for unit roots resulted negative, that means it is highly probable that VIX values are NOT a random walk, and as expected it holds a mean reverting nature.

2. Patterns, regularities around VIX returns.

VIX Day of the Week Effect (Keep in mind it's negatively correlated with the S&P500 stock index)

There's a statistically significant "Friday-effect".

This could be exploitable, that would take some further work. The bottom line here remains: financial time series do exhibit deterministic characteristics (ergo Random Walk theory sucks :).

2 Reflections:

Anonymous said...

you realize that the VIX isn't directly tradable, and the Friday effect is due to the repricing of SPX options for the weekend?

Rocko Chen said...

Yes you're right. I was thinking about VIX ETFs, futures, and options.

I didn't know about the weekend SPX option repricing. This makes it more interesting, then the statistical finding is even less likely spurious.