First Sort - Group by Country of domicile (it could be alternatively by Country Exchange)
The ranking first divides the entire universe into buckets based on their country of domicile.
Bucket 1: All U.S. Stocks
Bucket 2: All Japanese Stocks
Bucket 3: All German Stocks
... and so on for every country.
Second Sort - Group by Sector (Within Each Country)
Next, the ranking takes each country bucket and further subdivides it by economic sector.
U.S. Bucket -> U.S. Technology, U.S. Financials, U.S. Health Care, etc.
Japan Bucket -> Japan Technology, Japan Industrials, Japan Financials, etc.
Germany Bucket -> Germany Technology, Germany Industrials, etc.
Surprised there isn't more interest here.
A 'cheap' P/E in one country is often just the local baseline due to country risk.
Right now, my rankings are flooded with Finnish stocks likely because of the local economic downturn...
This is similar to another request we got from a professor for a way to do double ranking. He wanted to first group by quintiles based on a formula, then do a percentile rank in each quintile using a different formula (this is pretty common in papers).
I think we can kill two birds with one stone with this proposal:
We can have two "rank groups", with the first being a formula you enter. The second would be the existing industry & sector choices.
Use case 1
Here's an example for @pitmaster use-case for an EarnYield ranked in each country's sub-sector. The additional settings are circled in red.
The formula is just "CountryID" which returns an integer representation of a country (we don't have this now, but it's an easy add).
Use case 2
For the professor use-case the First Rank Group needs to return 5 distinct values for quintile groups. This can be done like this (using MktCap as example):
Trunc(FRank("mktcap",#all)-0.001)/20
It's a bit tricky bc the top rank is 100, so we need to subtract a tiny amount to get balanced 5 groups. Then he would pick "Universe" for the Second Rang Group. The end result is EarnYield ranked within each MktCap quintile.
Anytime we can satisfy two use-cases with one new feature, it automatically makes it's ROI much higher.
So there's no confusion: the double ranking groups in the ranking system is just brainstorming. It is not available yet, it's not finalized and has not been scheduled for development. Doesn't seem too hard to do.
Love that it would be formula based. So much flexibility and opportunities to be creative. Could even just do two formulas but have a dropdown with the standard choices
Another use case: Sort number of analysts and magnitude of earnings estimates revisions (or vice versa on the double sort). Rational: 5 analysts revising may be more significant that a single analyst revision.
This would capture something similar to (but not the same as) “Agreement’ used as a feature in Zacks Rank.
I think maybe this is generally better for capturing any interaction effects or moderator effects. Maybe we could ask the professor about the last.
But perhaps similar to “controlling for the number of analysts, magnitude remained significant” language we read in all those academic papers
