NAs

Research
1

when i run a ranking system, depending on the number of nodes, there are a number of NAs that show up for each stock. Can some one please help me understand the following:

  1. Is it possible to find out which items are NA for each security?
  2. What is the rational behind using NAs neutral scheme vs default and the relative advantages and disadvantages of each?

Thanks!

2

To answer your first question, if you run the Ranks page from the ranking system pages with the “Include Composite & Factor Ranks” selected for Columns then all the NAs will be highlighted in white. You can also look in the screener:

Sales(0,TTM)=NA

That will be true only for those companies with NA for sales in the last 12 months. (Hopefully, that’ll be a very small number.)

NAs negative throws the NAs to the bottom of the ranking. NAs neutral puts them around the median.

The primary reason to use NAs neutral rather than negative is that you care about the bottom of the ranking system. In practice, I would only say that this is mandatory when using the long/short option in the screener.

Personally, except in that one case, I use NAs negative exclusively. If I care about the both the top and bottom for some reason, I’d rather copy the ranking system and reverse the high-to-low/low-to-high output in the copy. That way I have total control of what’s happening.

3

One thing to note - if you have a 0% weight on a particular factor or folder I think the count of NAs shown on the Ranks page of the ranking system will include the 0% weight items as an NA in the tally. I don’t always clear out the 0% weights from a system and sometimes I’ll see high counts of NA due to this.

I mostly use NAs as neutral. For example, if a company doesn’t have an eps estimate for a factor and fires an NA due to that, I’m fine with putting it at a 50 percentile score rather than putting at the bottom of the rankings.

4

When designing a ranking system, it’s important to consider whether you’re going to have NAs be negative or neutral. For example, let’s say you’re using P/E. A huge number of stocks are NA if you use any conventional P/E measure because they have negative EPS, and if you put NAs neutral, those are all going to be in the middle of your rankings, ranked HIGHER than stocks with low EPS. If you’re going to have NAs be neutral, you want to avoid using certain factors. Since I want my NAs to be neutral, I use earnings yield instead of P/E.

Here’s where NAs being neutral can be useful. Let’s say you’re ranking stocks based on R&D spending. Certain industries (energy, banks) don’t expense R&D. So your formula needs to specify what to do with companies in those industries. Ideally, you’d want them somewhere in the middle. An energy company that doesn’t expense R&D is a better bet than a software company that spends next to nothing on R&D. The only way to put those companies in the middle is to use NAs neutral along with an Eval command (e.g. Eval(GICS(10,40)=1, NA, RandDTTM/SalesTTM)).

5

thanks Guys. i also noticed that spacemanJones. Yuval, great point, i have started to use yields too.