Monday, September 08, 2014

A New Measure of Daily Buying and Selling Pressure in the Stock Market

A while back I wrote about measuring buying and selling power in the stock market as separate variables.  This has been a project that has taken many months of trials and errors, not to mention countless spreadsheets.  The basic methodology is to disaggregate upticks and downticks for every transaction among NYSE stocks and then cumulate the upticks as a measure of buying pressure and the downticks as a measure of selling pressure.  The buying and selling pressure indexes are then set with zero as a mean.  High values indicate high levels of buying /selling; low values indicate low levels of buying/selling.

The backtests of the measures have been promising.  Going back to 2012, when buying pressure has been above zero, the next two days in SPY have averaged a gain of +.31%.  When buying pressure has been below zero the next two days in SPY have averaged a loss of -.09%.  As a rule, spikes in buying pressure have been followed by further index strength, suggesting that significant buying is important to upside momentum.

Also going back to 2012, when selling pressure has been above zero (above average levels of selling), the next day in SPY has averaged a loss of -.05%.  When selling pressure has been below zero, the next day in SPY has averaged a gain of .19%.  Interestingly, after that next day downside momentum, strong selling days tend to be followed by a bounce (average gain of +.44% over the next four days), while days with little selling show little subsequent price change (average gain of .07% over the next four sessions).  In short, spikes in selling tend to be followed by short-term downside (momentum) followed by subsequent recovery (reversal).

Daily levels of buying pressure correlate -.45 with daily levels of selling pressure, going back to 2012.  Buying and selling activity are thus not perfectly independent; nor are they wholly correlated.  Changes in buying (selling) pressure account for 20% of the variance in changes in selling (buying) pressure.  It is when buying (selling) pressure is unusually high or low for a given level of selling (buying) pressure that we see the most dramatic follow through in forward price movement.

Frankly this just skims the surface.  Aggregating the data over differing time frames and as a function of time of day yields quite promising findings.  Similarly, modeling markets as a joint function of buying and selling pressure increases the robustness of the backtests.  Much work remains to be done, as in-sample and out-of-sample backtests need to be validated by forward tests in real time.  I will be sharing data on the measures going forward and posting daily readings online if further tests bear fruit.

Further Reading:  Upticks and Downticks Across the Entire Stock Market