In response to many questions from the Short Selling Blog:

short selling

Short Algorithm

The screen that featured in the article can be described as follows:

·         US companies with market capitalisation greater than one billion dollars and with 20 day average trading volume greater than 100,000 shares.  These companies are less volatile than their smaller brethren and much more likely to be borrowable.  ADRs are excluded.   This gives a typical starting universe of about 1700 large, liquid US-based companies.

·         Then a 5 year earnings yield (EY5) is found for each.  The screen excludes those companies where with a 5 year average earnings yield  greater than 6%.  Remaining are companies with negative EY5’s and values belowe 6%.  Some may have attractive looking one year earnings yields but this can often be because of unusually high recent earnings.  This leaves about 450 of the 1700 companies.

·         Of those remaining, some may be valuable because of their assets rather than their earnings.  So all those with price / book ratios between 0 and 1 are excluded (or BM > 1) .  This leaves about 370 companies.  Some may have attractive looking P/B’s of, say, 1.01, and may feature in many alternative value screens.

·         All companies in the database are allocated to an industry sector.  In this database there are about 60 different industries.  Of the 370 companies remaining, I want to make sure that I pick maximum of one company per industry for diversification purposes.  I start out by picking the worst four in each industry from a capital control perspective.   I want companies that have been tapping the credit and equity markets for more finance.  These companies will have presumably been doing this because their cash generation is not strong enough or because the CEO wants to go on an ego enhancing acquisition spree.   This typically leaves about 150 companies, as many industries have no companies that meet the preceding steps.

·         Continuing on the diversification theme, the screen now picks the most expensive stock from the four in each industry.  For this, a combination value metric is used which allows the inclusion of loss making companies.   Now we are left with about 50 companies.

·         I always want a portfolio with the same number of stocks in it across the period of the backtest.  Otherwise, backtesting makes no sense as the backtester could pick a single company one period and 50 companies the next.  No one is going to invest like that in practice.  So for this article the screener picked the 12 most expensive of the 50 companies from 50 different industries.     To repeat: in each holding period, the portfolio consisted of 12 stocks from 12 different industries.

·         In practice, one might want to hold more.  I tend to hold 20 at any given time.

·         The last step could feasibly use momentum instead of value.  The results are similar.

I would like to repeat the caveats:  Actual results will not match those in the article backtest as not all the stocks will be borrowable in practice

.

However, because the screen starts with large, liquid companies and avoids momentum and the normal one-year value metrics, the stocks it picks are more likely to be borrowable than many short screens, in my opinion.  I have been trading this short strategy for several years and would estimate about 80% of the stocks it picks are shortable.  When I cannot short a selected stock, I pick the next stock on the list and often find little difference.

The other caveat is that despite assurances I have has to the contrary, there may be unknown biases I have introduced.  In which case, the backtesting results would be meaningless.   As ever with backtesting, one has to be happy with the logic behind the screen.  Ask yourself:  If I had not seen any backtest results, would these be the types of companies that I would not want to own?

 

Profitable investing,

Fraser Dawson