While reviewing the new P123 ranking systems, I noticed an issue that I’ve wanted to pursue for awhile; how does Stock Factor differ from Stock Formula when using *Ind factors? I thought they would be equivalent if the Stock Formula used Universe for the ‘Rank Formula Result vs Stocks in its’ selection, but I’m getting different rank values for each stock. It appears that somehow the Universe differs between the entries.
Also, when I change Stock Formula to rank within an Industry, it runs without error contrary to my expectations. Didn’t a combination of ranking *Ind factors within an Industry error-out before?
The Industry rank gives you around 76 different ranks (depending on the observation date), one rank per each industry in GICS, with a delta of 100/76=1.3 , evenly spread since it does not matter how many constituents there are in each industry. The only thing that matter is how many different industry values there are and how many NAs which get bunched together at the bottom.
The Universe formula with the Industry Factor inside is the closest to the above. You will still have around 76 different values but in this case the size of the industry matters : the delta for the next value will be proportional to the size of the industry
The Industry formula with an Industry factor inside makes no sense . There’s only one value for each industry so each stock will get around the highest score for that node.
So cap-weighting is involved? Is that what you mean by size of the industry?
I know! That combination use to error out, didn’t it? Better yet, why doesn’t it generate an error now?
EDIT: The Industry formula with an Industry factor returns ranks from 99.90 to 1.20 (for the S&P500 Universe). So it’s ranking something across a large set.
By size of industry I mean the number of stocks in the industry. If all values are different every stocks’s rank differs by a percentile step defined by 100/#stocks. When values are equal the percentile step is multiplied by the number of stocks with the same value.
Not sure it ever generated an error. The computer code to “understand” that this combination is meaningless would not be trivial.
As for the error report, I’m probably mistaken. But does the code need to understand anything other than *Ind factors can’t be used to rank in a Industry? Anyway, that’s above my pay grade and I won’t worry about it.
