New post out of CXO Advisory, testing certain indicators to see if there is any predictive power in the readings. Their analysis of Put to Call ratios show that there is very little correlations (either positive or negative) with respect to any ability to time the market. Their conclusions:
In summary, evidence from simple tests is too weak and inconsistent to support a belief that stock market put-call ratios reliably predict future market returns.
Bad Sampling
So it seems that another indicator bites the dust, right? Well, not so fast. Let's look at a chart of the CPC, provided by stockcharts.com:
Just looking at this chart over the past year shows that it's very, very, noisy. Actually trying to attain signal out of every single data point would be a fool's game. Therefore, testing a 1-1 correlation of put/call readings and returns on the SPX is not the best study to determine whether this indicator is viable or not.
With noisy charts like this, I generally look at two things: the extremes and the distribution. It's the same thing with the NYSE TICK charts: anything inbetween +/-800 is pretty much noise and dosen't reflect the supply and demand of the market.
Test the edges
The same thing happens here: any readings not extreme are probably noise and don't reflect the sentiment of the market. Instead, I propose a new set of tests.
- Look at correlations when put call ratios hit a specific extreme (maybe y<.7 and y>1.15). This will show whether the "blood in the streets" trade works when a bunch of traders buy puts.
- Look at correlations relative to a smoothing mechanism, most likely some moving average, that way the data gets less "noisy" -- this may increase signal but reduce hit rate