Wednesday, April 8, 2009

Is SPX Down Big & VXO Down Bearish?

Interesting about today is that while the S&P dropped well over 2%, the VIX (and VXO) also dropped. Often traders will interpret this as bearish. The thought process is that when the VXO doesn’t drop with the market it suggests a possible complacency among traders. The theory is that this complacency may lead to further selling until participants become somewhat fearful again. At that point a rally will become more probable.

I looked at this theory last month when considering action over the course of several days. At the time I found no evidence to support the bearish case. For testing I looked at a 5-day divergence to measure the return in the VXO and S&P prior to triggering. Tonight I looked at a 1-day divergence.

Not only is there a lack of bearish evidence but these results could be considered borderline bullish. I also looked at more extreme 2% SPX drops:

If the number of instances wasn’t so small then I’d certainly consider the results strongly bullish.
A sensible sounding theory is just that – theory. This is an example of why I make every effort to test all of my trading and market bias ideas.

Note: Apologies for the sporadic posts this week. Things should return to normal next week.


Douglas said...

Thanks for this. I test some of my theories also and get counter-intuitive results.

I tend to look at what implied volatility is doing over time and I would say that something like this adds to a longer term bearish case even though the one-off events on their own can have an effect that is the opposite...

Anonymous said...

Thank you also for the test. Would we not have expected a more bullish resolution as the exact opposite (both SPX and VIX raising together) is bearish? - Thus if both fall in concerto it should be bullish?

Thanks, Joe

Shawn said...

Rob, thanks for the free daily upates. Here's my 2 cents on VIX:

1) VIX or VXO is the leader. shh... this is a secret. So when market down with VIX is down, it means bullish for a market reversal soon (VIX down implying stock should reverse to up soon). Vice versa.

2) However VIX is stuck in a big triangle lately. Draw the trend line from High of July 2008, multiple lows in Dec 2008 and Jan 2009, Lows in March. That's a support.

Now draw the resistance from High in Nov 2008, through highs in Jan and March, you will see a big triangle. It got to resolve soon.

However, there was a recent post on Zero Hedge that VIX need to be interpreted in conjunction with sovereign CDS as more countries are shifting banks/financials toxic risk onto sovereign books.

Those countries CDS have jumped a lot lately. In brief, smart money have partially move to a different playground, rendering VIX alone not a good predictive tool.

Douglas said...


Inspired by the way you crunch price data I decided to have a go at an analysis of what has happened in the past after very large % increases in the market over the period of a month.

(I build excel models to crunch the numbers but normally do very different stuff to your work).

I have daily stock market (close price) data going back to 1928. I simply put in a formula to see what, for every day since 1928, the % increase in the market was over 22 trading days before (this may not be perfect as I understand that in the 'olden days' they traded 6 not 5 days a week).

Then I used some If formulas to find big increases - of course to get more than results in the 1930s I had to lower the bar as there has been nothing like the past month or so since 1938.

I then looked at how the market had done % wise out to 22 trading days later.

The rule was that once the 22 day increase had dropped below the specified % the following 22 days would be looked at.

I started by looking at all instances where the market had risen 14% on where it had been 22 days earlier. Then I increased this by 1% each time.

On the face of it the results looked bearish (I calculated the average, max, min and median for each of the following 22 days).

However, I think that my model is flawed in that if a surge followed within 22 days of an earlier surge the calculation on the first surge would cut at out that point and restart counting from the new one.

I did all this with excel functions rather than vba (because it is late at night) so I was not able to deal with this scenario.

Have you thought of looking at instances where the market has gone up a large x% over 22 trading days and then what happens next? Your models are probably set up to do this while mine are not.


PS I looked at two sets of results - all data and all data excluding the 1920s and 1930s.

PPS Looking at your earlier posts this looks similar to some of your earlier posts on new highs under the 200 day MA. However, this is a bit different as it is about a month long sharp rise with not much of a breather on the way. The only time I can see, at a glance, when a rally went on longer and harder and sharper was from the seemingly absurd levels in the summer of 1932 - otherwise it looks to me like we must be due a significant drop even if this is only a rest on the way to higher levels...