Bernstein Seasonality and the PSEi

Actually it’s probably not really called Bernstein Seasonality, but I first encountered the methodology from his book on seasonality in futures markets. Interestingly, I also encountered the same methodology being used by a trader I encountered in the Finance Manila forum by the name of choobeebo, who is a futures commodities trader who has serviced some big name clients (San Miguel Corporation to name one).

So if it’s such a popular method, I thought why not try it on the PSEi?

The concept behind Bernstein Seasonality is to normalize yearly price movements to discern likely price action during certain months of the year. I’ll detail the basic steps of the method along with the data and results as well.

To start, since we’re going to be testing monthly seasonality, you have to get the average price for each month of the year in your data. I did this by getting an average price for each trading day–which is the average of the highs and lows for the day (which I think is more representative of the activity, rather than the close).

Once I got the average daily prices, I got the average of each month and laid them out on a chart like this:

Now to add flavor to the later part of the analysis, I chose to label the years depending on their trends. For each year, I got the average price for the year (simply the straight average of all months yearly). Then for simplicity, my measure of trend is simply the December average price. If it was lower than the yearly average, then it was a bear year. If it was higher than the average for the year, then it was a bull year.

The result is fairly accurate: 1999-2002 were bear years, while 2003-2007 were bull years. 2008, while not complete, I chose to label as a bear year, firstly because the average June prices were lower than the annual average, and secondly because of the events surrounding 2008.

The next step is to normalize the prices by dividing each month’s average price by the yearly average. This would reduce the actual values into integers that indicate the magnitude of those monthly prices in relation to each other:

Now we have the material we need to compare each month with each other. If we graph each normalized year from this table, with our months as data points, here is the result:

Interesting thing here is how the bull and bear years tend to cluster together. Also note the position of our current “bear” year 2008 in relation to the rest of the pack. Pretty bad eh?

First thing we can deduce from this graph is that although each cluster has its own tendencies, August seems to be a pivotal period which reverses the tendency. So if a year was good prior to August, the remainder of the months would be bad, and vice-versa.

Now to smoothen out the dispersion, let’s group the bulls and bears together and compare them:

Don’t mind the colors: the RED line are the bull years, while the BLUE line are the bear years. Some observations here:

• For Bull Years: Dip in March, rally to July, then dip to August, then rally to December.
• For Bear Years: Dip in March, rally to April, Dip to June, rally to July, then dip to October, rally to December.

Of course, we only know a year as bullish or bearish, pretty much after the fact. So what can seasonality tell us in general? To do this, we can merge all years into a grand composite to see what kinds of tendencies each month has in ALL years:

For all years: Dip in March, rally to July, dip in August, and pretty much stagnant till October, then rally in December which continues to February.

Insight and Application

The Bernstein Seasonality seems to support the earlier results of the Williams TDW study and the original Seasonality Problem on the insight of October being a crucial month. However the directional behaviours of March, July, and December are highlighted here, which prevail across pretty much the last bull and bear cycles of the PSEi.

Again, like the previous seasonality studies, the results here are not to be taken as mandatory–but can serve as support to existing technical and fundamental analysis in formulating investment decisions. The main limitation of technical input is that it is largely reactive in application, while fundamentals only give you valuation targets, but not timing strategies. These gaps can filled in by seasonality studies.

Maybe if you roll everything into one strategy:

• Fundamentals provide your valuation targets, economic bias, and stock selection.
• Seasonality provides your timing strategy.
• Technical Analysis provides your entry and exit criteria.
• The last piece of the jigsaw is your money management to tell you how much to bet.