Thursday, December 07, 2006

A Different Way To Measure Market Sentiment With Options Data

I recently mentioned that I had begun a large research project to investigate historical patterns connecting options data to future movements in stocks. Some of my initial efforts have focused on the equity put/call ratio and its relationship to future equity index movements.

Suppose, however, we treat the put option data and the call option data as completely different time series. Might there be information in each that is obscured when we solely focus on the put/call ratio?

As a result, I calculated the ratio of each day's equity put volume relative to the average equity put volume over the past 100 trading sessions. This we will call the put ratio. I then calculated the equity call volume relative to the average equity call volume over the past 100 sessions. This we'll refer to as the call ratio.

Armed with a separate put ratio and call ratio, we can address the question of whether relative peaks and valleys in equity call and put volume reflect sentiment shifts that impact future stock prices.

Additionally, we can form a new relative put/call ratio that is the put ratio divided by the call ratio. In other words, the relative put/call ratio identifies whether we're getting more put or call activity in the market as a function of the past 100 days' volume.

It turns out that elevations in puts and calls have different impacts upon the S&P 500 Index (SPY). And the relative put/call ratio does a pretty good job as a sentiment measure and as a predictor of S&P 500 prices 1-20 days out. Tomorrow I will begin detailing some of these findings.

For now, suffice it to say that the relative put/call ratio got very low in mid November prior to the short, steep market drop. We have since moved to neutral levels in the put ratio, call ratio, and the relative put/call ratio.

When we have had such neutral levels on a one-day basis (N = 126), there has been no overall edge--bullish or bearish--over the next 1-2 days. When the neutral day occurs during a bearish five-day relative put/call period, however, the next two days in SPY average a gain of .24% (40 up, 23 down). When the neutral day occurs during a bullish five-day relative put/call period--which is what we have at present--the next two days in SPY average a loss of -.11% (27 up, 36 down).

The new options measures clearly point out that returns are superior following five-day periods of bearish sentiment and subnormal following five-day periods of bullish sentiment. By measuring bearishness (put ratio) separately from bullishness (call ratio), we can tease apart periods in which the traditional put/call ratio is high because of high put activity vs. low call activity. More soon to come...

11 comments:

Anonymous said...

Keep it coming Brett!

John Wheatcroft said...

Even though options are used to address future expectations they are bought and sold based on past experiences i.e. "market down - buy call/sell put" "market up - buy put/sell call" or an infinite variety of riffs in a similar theme. And as you know of course, the whole subject is lousy with the psychology of trading or maybe the "psycho" nature of trading - "good earnings - sell stock; bad earnings - buy stock" which is contrarian to a fault but remains a profitable trade and everyone knows it. And is many times done with options rather than actual stock.

I'm thinking that the correlation factors in the market (stock to stock, beta to beta) are extremely low at this time. That suggests that a lot of randomized buying and selling is going on which is keeping overall volatility (VIX/VXN) at a decreased level. (A function of buying the dips and selling the rips at the micro level repeated maximally). Couldn't that also have a telling effect on the numbers of puts and calls being sold or bought? In other words is it possible that your research is being informed by market dynamics and your findings are only coincidental to the environment?

Or, better, how do you normalize your findings to eliminate possibility of skew?

Also since put/call activity is generally keyed to the next 20 - 30 days or so why not use a shorter average for normalization purposes - such as 20 days?

Very interested in this subject - thankyou for your time and effort.

Brett Steenbarger, Ph.D. said...

Thanks, Ben; there's a lot to study when it comes to options!

Brett

Brett Steenbarger, Ph.D. said...

Hi John,

You're raising good questions. One is whether or not the options data (which does have some correlation to past price change) *uniquely* predicts future price movements. The second issue is whether we might see different and better results by looking at options volume as a % of the last 20 (rather than 100) days. I would add a third issue, which is index options and their elevation of volume. I hope to get to all of these in the not-too-distant future. Thanks--

Brett

Anonymous said...

Brett, Have you considered looking at option open interest? It should be possible to aggregate the delta and gamma of open positions on all options for a particular stock/index and work out a NET position for participants in that underlying. For example you might be able to see that the options market is net short delta in X size and that the gamma suggests the option market needs to buy Y shares on every Z dip to stay flat?

Just a thought.

Q1

Brett Steenbarger, Ph.D. said...

Hi Q1,

Those are great ideas. What source do you use for historical options data? Thanks much for the worthwhile direction--

Brett

Anonymous said...

Just wanted to add to John's comments that the options usually look ahead 20-30 days.

I believe options with 30 days to expiry are trying to predict the underlying's 30-day HV.

That's why the VIX measurement is tweaked / normalized to a 30-day, ATM option.(Read that in a blog somewhere yesterday)

Similarly, options with less time to expiry (eg 10 days) are trying to predict 10-day HV, as I understand it from my research on the topic.

What's my research? I predict stock option IV after earnings announcements, using a history of IV and HV and I'm getting accurate within 2 percent, though options with less than 14 days to expiry catch me out occasionally.

Cheers / Ben

Brett Steenbarger, Ph.D. said...

*Very* interesting research, Ben; thanks for the comment and the perspective on time horizons--

Brett

Anonymous said...

Actually don't use this myself (blush) but its been on the "i must look into that" list for quite a while...

Q1

Brett Steenbarger, Ph.D. said...

Hi Q1,

I do think it's worth the investigation. Next I'll look at index options. I'm a little more skeptical of their value, given how often they're used as hedging vehicles. But we'll see! Thanks--

Brett

Noel & Kaye said...
This comment has been removed by the author.