FactSet, GICS, and RBICS

We will shortly be releasing a beta site which will give you a chance to play with FactSet data.

FactSet classifies companies differently from S&P: it uses a system called RBICS rather than GICS. I have prepared three documents which will help you during this transition. The first is a document explaining the similarities and differences and how to manage the transition. The second and third are Excel files, one with mapping data, the other that lays out the RBICS structure, including all the subindustries. It’s very important for all users to read the Word document at some point. Don’t worry–it’s only three pages long.

The key points are:

  1. GICS and RBICS are similar in that they both use a four-level hierarchy (sector, subsector, industry, subindustry) with two-digit numbers divisible by five assigned to each.
  2. Most of the sectors are the same.
  3. There are two RBICS sectors that are not GICS sectors: business services and consumer services.
  4. There is one GICS sector that is not an RBICS sector: real estate.
  5. Individual companies in a particular industry will almost always be different between GICS and RBICS. A company that ranked highly in its industry according to GICS may find itself among rather different constituents in RBICS and that will affect its ranking.
  6. All of your existing commands will still work and you won’t have to change anything. We will be mapping all GICS codes to RBICS codes. The mapping, however, is necessarily imprecise and as time goes on I strongly suggest you abandon GICS and work with RBICS instead.
  7. We have five new commands available for your use on the beta site: RBICS, RBICS.Sector, RBICS.Subsector, RBICS.Industry, and RBICS.Subindustry. Instructions on how to use them are in the Word file.

Please reply to this thread if you have any questions.

Google links:
Word document: https://drive.google.com/open?id=1moiYxmCkHlqEhgLCcQoZwiWRb5rBiOvs
GICS to RBICS mapping: https://drive.google.com/open?id=1UQ4xrncttO7FtFtQXkBbvvJglprrik6r
RBICS structure: https://drive.google.com/open?id=1-6A_7aSsshv0gudWLB3M7AeSCZlXOak6

I suggest you open/save the second and third links as Excel files.

Will FRank and similar functions work on RBICS or GICS classifications for #Industry and #Sector at the beta site?

Thanks,

Jeff

I second Jeff’s question, and I extend it to all cross-sectional filters:
#Sector
#Industry
SecCount
IndCount
SecWeight
IndWeight
Now, these filters are based on the first 2 digits of the GICS for “sector” and 4 digits for “industry”.
Will you:
A- change them to the first 2 (resp 4) digits of RBICS? (in this case our rules will drastically change behaviors overnight)
or
B- keep the legacy cross-sectional filters thanks to your mapping table and add new cross-sectional filters based on RBICS? (in this case our cross-sectional rules based on GICS will keep almost the same behavior, and we can develop new ones based on RBICS)

For the beta site, FRank will continue to work only on GICS industries, sectors, etc. SecCount, IndCount, SecWeight, and so on will also work only on GICS. The beta site will evolve as we get more and more FactSet stuff on it.

The filters are based on the first two digits of GICS for sector and first six digits for industry. Once we switch to FactSet, the same will be true. When we switch over completely, no GICS information about individual securities will be retained in our database. GICS codes that are in your rules and formulas will be mapped to the appropriate RBICS codes, but IndWeight, SecWeight, and so on will all necessarily be based on RBICS classifications.

What do you think these Saas software companies (OKTA COUPA TWLO MDB ESTC PLAN) will end at, in terms of Operating margins?
Even the bigger ones WDAY CRM NOW aren’t really profitable. How do you estimate the margins?

Like MSFT has huge margins so do you assume all these negative margin companies get MSFT margins?

Hi Yuval,

Given your previous investigations into company classification, your continuing insights will be welcome.

Questions…

When the beta site goes live, will it be fully populated by the FactSet data we will have (and need to rely upon) come July 1st or will there be additional FactSet (or other data provider) input layered into the beta site during the transition period?

Likewise, will the current CompuStat data-driven site be left completely as is for the duration of the transition?

I ask (and I’m sure I speak for many subscribers) because we would like to feel certain we can proceed on a parallel basis until June 30th by relying on existing screens and models as is while getting up to speed with exactly the data - no more, no less - that will populate our screens and models beginning July 1st.

If I had all the time in the world, it would probably be interesting to see how RBICS performs with CompuStat fundamental data. Or GICS with FactSet data. Is that going to be possible, by the way? As a practical matter, I believe I am going to have to concentrate on running RBICS with FactSet data. Mixing and matching the two data sources, if possible, will be for purposes of exploring unexpected - perhaps I should say unfavorable - results.

Also, will it be possible to simulate daily trading systems on the beta site?

Finally, is July 1st the actual date that the CompuStat data will no longer be available?

Thank you.

Hugh

https://insight.factset.com/resources/factset-revere-business-industry-classifications-datafeed

[i]There are four packages available in the RBICS suite:

RBICS – Focus
includes single-sector mappings of approximately 48,000 of the most-liquid and publicly-traded companies based on their primary lines of business.
RBICS – Extended Universe
includes single-sector mappings for approximately 3 million private and public companies, on a current-only basis, at Level 4 of the RBICS taxonomy.
RBICS – Revenue
includes a multi-sector mapping of approximately 45,000 of the most-liquid and publicly-traded companies. Revenue percentages associated with each reported business segment are standardized into the most granular sectors of the RBICS taxonomy.
RBICS – Tradenames
includes 170,000 products/services (Tradenames) associated to each line of business within the RBICS company universe.[/i]

Any idea to which packages P123 and its users have access?

Thanks for your answer Yuval. So I understand it is impossible to test RBICS on the beta site with screens/sims using SecWeight, SecCount… or using cross-sectional functions (FRank, FOrder, FMedian, etc… with a parameter ‘#sector’ or ‘#industry’) . I guess it is the same for rank nodes using industry factors. If it is correct, we’ll have no clue on the new behavior of these screens/sims/ranks until the day of switching.

As I said, the beta site will evolve. At first, it’ll only have fundamental factors and RBICS codes (these are for reference only, not affecting ranking and grouping) and it’ll be based on the stocks in the Compustat database. Marco’s in charge of this project, so he’ll be able to tell you in what order things will be added to the beta site. At some point, the beta site will offer both the Compustat version and the full FactSet version of Portfolio123, with all factors and all FactSet stocks and all the RBICS stuff fully implemented. But we’re not there yet. It’s an enormous project.

Portfolio123 and its users will have access to levels one through four of the RBICS taxonomy, both current and historical, for all stocks in the FactSet database. This use case doesn’t conform to the above package descriptions.

Yuval, thanks again for your answer.

I think the Beta site should contains only the FactSet data, same as will be when the transition complete. All existing functions should run solely on FactSet. This way we will be able to see the differences and amend our ranking and systems accordingly. Please don’t ask us to change all of our systems with new functions.

Can you clarify something - will the existing commands for “industry” searches still work the same? In other words if we use “industry” as the search term, (and not GICS) how different will the results be compared to S&P data?

On the beta site, commands like “industry” and “sector” will refer to the GICS classifications and “RBICS.Industry” and “RBICS.Sector” will refer to RBICS classifications. That way you can compare the two sets of classifications.

But that means that we need to change all our systems in order to check the new data (change “industry” to “RBICS.Industry” ect.). Why not to compare between the beta site results (which “industry” refers only to RBICS) to the current site (which “industry” refers to BICS)? This way we will be able to know where are the differences and where to put are efforts to.

Yes this is exactly what I’m wondering. I see the earlier comment that all of our existing commands will work - but is that correct? Will “industry” searches need to be modified?

Because if we did that, your formulas would be REALLY messed up.

Let’s say, for example, that you have a universe rule that reads Sector != 10 because you want to avoid buying energy stocks.

We are going to automatically translate that, internally, to RBICS.Sector != 25.

If we were to do as you suggest and switch the code to RBICS, you would end up excluding all industries in the business services sector.

It’s really up to you whether you want to modify them or not. Let’s say you’re looking for companies in Technology Services. If you use Industry = 451020 (IT Services) you’ll end up with companies in RBICS industries 101015 (except a few), 552020, and about half of those in 552010. On the other hand, you could use RBICS.Industry = 552010 or RBICS.Industry = 552020 (Internet and Data Services and Technology Consulting Services) and come up with quite a different set of companies.

The main goal here is that users won’t HAVE to modify anything. Everything will still work because we’re mapping all GICS codes to RBICS.

Mostly I use text searches, not GICS codes, as in Industry = SOFTW.

We’ve assigned names to RBICS codes too and as much as possible they’re the same as the GICS names. Industry = SOFTW will give you RBICS 552015 (except 55201520), 20151530, 30301510, 60101035, and 60101040. RBICS.Industry = SOFTW will give you simply RBICS 552015. I’m not sure if the industry names will be on the beta site at first, but if not they’ll be there quite soon.