MORE TECHNIQUES --> | Investing Best Practices |Reducing Strategy Risk | How to use site|

Since mechanical Strategy Indexing relies heavily on fundamental metrics taken from companies' financial reports, it would stand to reason that the age of the data we use for making investment decisions is an extremely important factor. At PowerStocks Research, we have a unique approach to examining fundamental metrics (such as P/E, ROC, Price/Sales etc.)  as opposed to traditional international research methods. When examining shares' financial reports for strategy back-testing or portfolio screening/rebalancing purposes, we look at the MOST RECENT financials (be they interims of finals) as opposed to just finals as is traditionally the norm. This ensures we are using the "freshest" fundamental data possible.

This is extremely important, especially for automated stock picking and screening purposes, since you may run the risk of making decisions based on very old valuation data that no longer matches reality, leading to bad investment decisions. This is especially true if one considers that a company has a 1-3 month lag between when its financial period ends and when it actually publishes that information.

By way of extreme example, assuming you only screen using the last final results, then some shares in your screen may very well have up to 15 months old data! This would occur if a companies' final ends on 30 August 2007, they publish it on 30 November 2007, within the JSE 3-month deadline, and you are running a screen on 29th November 2008. But if we were using interims as well, then the above example would result in us using the 28 February 2008 interim report which would only be 8 months old.

We saw that by using interims as well as finals in our screens and tests, we can significantly reduce the average age of our data. But we can leverage another characteristic of the JSE to our advantage to significantly freshen our data even further, namely the earnings season profile. The below graphic depicts the financial period-ends of all listed JSE companies by month.

We see that there are two major "earnings seasons", namely June and December with roughly 205 companies reporting a final or interim, and 4 minor "earnings seasons" in February, March, August and September, with roughly 80 companies reporting results in each month. The profile above was taken from 2008 data, but we have examined the profiles going back to 1999 and there is a less than 1% standard deviation, so you can safely assume this profile is cast in stone. It stands to reason that by running your backtests and screens on dates optimised around these "earnings seasons", you should be able to significantly impact the age of data being used.

But what we have to remember is the lag between period-end and publication date, which effectively shifts the above periods by between 1-3 months. You will notice that all our backtests on this site have picked a start date of 31 March. This is because using the JSE reporting deadline rules, we can safely assume that all the companies will have published a final or interim for the previous calendar year on this date. Also, shares in the major December earnings season (whose periods end on 30 December) will have published their results by 31 March, and we will have a bunch of fresh data. As an additional bonus we pick up some early results announcements from the February/March minor "season" as well (some companies report within 4-6 weeks). Similarly, 30 September would also be a good date to pick for portfolio scanning, but not for back-testing as you would be picking up results straddling 2 calendar years. March ensures our backtests are all based on "very fresh" data of all companies whose results fall within the previous calendar year.

We have performed a detailed analysis for each month of the year, to compute the average age of fundamental data taken from the most recent published final or interim, to figure out the best times of the year to rebalance or create portfolios. The results are below:

Firstly, we see what a MASSIVE impact using interims can have in reducing the age of fundamental data. Using Finals only, the best you would hope to achieve is in the month of August with 4 months average age from publication date (6 months on average from period-end date). But by switching to Interims OR Finals in our selection criteria, we reduce this to 1.8 months!

Secondly, we see that February, followed closely by August are the best times to create/rebalance portfolios, since these dates reduces the average age of your fundamental data being used in screens to an astonishing 1.5 months from publication date (3.5 months from period-end date). May and November are not too bad either, at 2 months average from publication date (4 months average from period-end date.)

January, July and December are not good dates around which to pivot mechanical stock picking strategies since they have data that is an average 4 months old from publication date (6 months from period-end date.)

We have scheduled a very detailed research paper on the effects of using freshly ground fundamental data versus "Finals" data, across various strategies and inception dates, but to satify our own curiosity, we ran backtests on two "Low Price/Book Decile" strategies, one based on a screen of Finals only, and another on a screen of Finals+Interims. Over a 5 year period, with annual rebalancing each 30th March, the strategy using freshly ground fundamentals returned 738% versus only 448% for the strategy using "stale" data. That is a massive difference!
Make a Free Website with Yola.