Seasonal vol patterns?

aid1961 · Jun 13, 2005

Just brainstorming here...is anybody trading vol on sectors in a seasonal manner, ie like seasonal spreads in futures but without the pair-trading aspect. I haven't done much research on it yet, but it seems that vol on retail stocks for example should rise ahead of christmas sales, energy stock vol during the winter, biotech stocks ahead of all those summer conferences, etc. Anybody have some insight on whether patterns like that exist and are reliable?

neural network software coupled with IV time series could probably find some interesting stuff.

Trajan · Jun 15, 2005

Yes, I've thought about, but haven't run anything yet see if something is there. I guess that it would show up in the partial autocorrelation function graph, but I'll have to think about it some more.

edit: after loading the data, I discovered that I have no desire to program this up right now to do it right.

murray t turtle · Jun 16, 2005

Partials Quote from aid1961:

Just brainstorming here...is anybody trading vol on sectors in a seasonal manner.

I haven't done much research on it yet, but it seems that vol on retail stocks for example should rise ahead of christmas sales, energy stock vol during the winter...
Anybody have some insight on whether patterns like that exist and are reliable?

neural network software coupled with IV time series could probably find some interesting stuff.

=======
Even though thats a strange name ,''retail'' ,couldnt help but notice JCP sector has done pretty well ever since Carl Ichan bought in, years ago. Mr Penny named his boy's middle name ''Cash''

Bid ask spread was better than i thought for options;
but actually havent had much time to research sector much either.

aid1961 · Jun 17, 2005

Trajan, please keep me up on what you find. Also, what software are you using to test this?

Aside from using neural networks, it could be done by charting an average of implied vols for various sectors across 10 years or so. The chart would look like a one year chart. The x-axis would have months jan through dec, and the y-axis would have the sum of the 10 years of implied vol divided by 10.

Does that approach make sense to you? If you had a chart like that, you could make a conclusion like "implied vol on drug stocks tends to rise in july and get smoked in september" or whatever.

Let me know what you think.

Seasonalcharts.com has some seasonal SV charts for a few big names, which are quite interesting to tell you the truth, although it's pretty limited what they have there.

Trajan · Jun 17, 2005

Matlab and Eviews, the latter is easy to use for time series, but the former allows more versatility. I had loaded some data in Eviews only to realize I probably need Matlab for this. Data manipulation is a little easier in matlab. It was when I had realized that I needed to calculate each month is when I quit, but I could probably do it in short order in matlab.

I think that would tell you something, but I would say that you would need to watch for outliers. So look at the underlying data. Also, the median probably should be looked at for this reason. Even if one or two outliers cause the rise, it could still be a good trade since events around that time period could repeat.

I'm going to review some stuff on seasonality

aid1961 · Jun 18, 2005

Trajan, that's a good point about outliers, and taking the median probability is probably a good fix. Would you be willing to share your findings via private message or publicly, or would you rather not?

Trajan · Jun 18, 2005

Quote from aid1961:

Trajan, that's a good point about outliers, and taking the median probability is probably a good fix. Would you be willing to share your findings via private message or publicly, or would you rather not?

Sure, I'll post or send a pm.

I found a paper which lools at the Swedish stock markets seasonal volatility patterns. Here. It has some interesting ideas on how to test this.

This not entirely unrelated quote is interesting:

One remarkable feature is that average stock returns are highest in April and May at the same time as the standard deviation is quite low! Return per unit of risk (coefficient of variation) for the two above-mentioned months are 0.16 and 0.29, respectively. Even February and June show figures on a level with Aprilâs return per unit of risk. One explanation for the high returns for April and May is that dividend payments are concentrated to these two months.

aid1961 · Jun 19, 2005

Trajan, I actually came across that paper as well...pretty interesting. Looking forward to seeing what you find.

stephencrowley · Jun 19, 2005

Seasonal volatility is not complex enough to require neural networks.

An Introduction to High-Frequency Finance is a book I recommend.

Quote from aid1961:

Trajan, please keep me up on what you find. Also, what software are you using to test this?

Aside from using neural networks, it could be done by charting an average of implied vols for various sectors across 10 years or so. The chart would look like a one year chart. The x-axis would have months jan through dec, and the y-axis would have the sum of the 10 years of implied vol divided by 10.

Does that approach make sense to you? If you had a chart like that, you could make a conclusion like "implied vol on drug stocks tends to rise in july and get smoked in september" or whatever.

Let me know what you think.

Seasonalcharts.com has some seasonal SV charts for a few big names, which are quite interesting to tell you the truth, although it's pretty limited what they have there.

Trajan · Jun 19, 2005

Quote from stephencrowley:

Seasonal volatility is not complex enough to require neural networks.

An Introduction to High-Frequency Finance is a book I recommend.

I've been meaning to get that book and I see that it directly addresses the issue. Anyways, I was actually away for the weekend, but I'm back and will get started on this. Maybe I'll search some of the finance Journals and see what's there.