Earlier this week, I took a look at the S&P 500's correlation to a number of less-frequently-cited indicators and got some interesting results.  Though far from perfect conceptually and even statistically, they did point to an expected negative trend for the coming month.  The indicators I had used were the put-call ratio, which is simply the total volume of puts traded divided by the total volume of calls, the short-term trading index, and the well-known CBOE volatility index VIX. 

    While working with this data, I decided to look for cycical behavior in the correlations.  Here's what I did:

• Calculate the return correlation of the put-call ratio, short-term trading index,and VIX with the S&P 500 over the last 5 sessions.
• Calculate the time series representing the average of each of these indicators at each time.
• On this average time series, calculate the historical 2% and 98% percentiles, that is, the correlations at which only 2% are lower or higher.
• Find all dates on which the closing average was in this extreme 2% case.
• For dates in the lowest 2%, pick only the dates on which the S&P 500 closed below its 10-session moving average.
• For dates in the highest 2%, pick only the dates on which the S&P 500 closed above its 10-session moving average.

    Now, what we have is a historical buy-sell signal set.  For each buy date, purchase shares of the S&P 500 ETF tracker SPY, and sell them next time you get a sell signal.

Here's what the average indicator chart looks like after smoothing (it's very noisy).  Click through the image to see the large version.

    Of course, these percentile and moving-average parameters can all be tuned infinitely, but as an example, here are the resulting signals for the numbers above.  The average return is 13%, beating the annualized S&P 500 return since 1990 of 11%.  Not only that, but the average hold length was 172 sessions - less than 70% of the time a buy-and-hold strategy would leave your cash at risk.  Though there are obvious tax ramifications, the redution in risk is significant, and easily mitigated if one were to hold yield-bearing assets when not invested in the market.  This easily beats the cyclical premium/discount strategy I discussed earlier.

    One interesting thing to note, however, is that given the recent changes in the systemic market volatility levels, no signals have satisfied the system's criteria in 2007.  Obviously one could look only at the recent history of these indicators to obtain percentiles more efficient for the current market, but given its reliable past performance and risk profile, I personally am quite happy to err on the side of caution.