[Feature]: Add characteristics from prominent asset pricing models

[theisij]

Feature Description

For ``the power of the common task framework,'' we used several characteristics related to prominent asset pricing models, which @faheemalmas added to the SAS version of the database. Now, we need to implement these characteristics to the Python version of the database.

Use Case / Motivation

Get coverage of exact factor characteristics used in prominent factor models. For example, our be_me characteristics is not directly comparable to Fama-French's book-to-market characteristics, so we need to add a be_me_ff3 characterisic (this is also related to issue #33, which should be closed once we complete this change)

Academic / Industry Justification (if applicable)

No response

Proposed Implementation (Optional)

No response

Willingness to Contribute

No, but I can provide detailed specifications

Alternatives Considered

No response

Additional Context

@FernandoRDLL: please see the email that I sent from Faheem, with all the neccesary SAS code. You job is basically to take what Faheem did in SAS, and integrate into our Python database. Once it's implemented, we should check our factors against the corresponding factors from the authors website. We can talk more about this as we get close

Pre-submission Checklist

• I have searched existing issues to ensure this feature has not already been requested
• I have read the documentation to confirm this feature does not already exist

[FernandoRDLL]

Ok @theisij, should I also add the portfolio code?
That one should be easier because it's already in Python

[theisij]

Yes, definitely. We should implement the creation of the asset pricing factor models as a separate module in our codebase.

Actually, all of this should arguably replace the current implementation that we have in ap_factor(). Specifically, this part is, as far as I remember, a quite ad hoc implementation of the FF3 and HXZ4 models, with whatever characteristic we had at our disposal (so, e.g., be_me instead of be_me_ff3). We should replace those factors with whatever Faheem has created, and then those asset-pricing factors become a key output of our codebase, and they will also be used downstream for characteristics like ivol_ff3_21d.

If you think all of this should be broken up into multiple issues, feel free to do so.

Does all this make sense?

[theisij]

Basically, this issue has two layers:

1. Create characteristics that closely mirror the construction in the original asset pricing factor paper (e.g., create be_me_ff3 such that it closely mirrors Fama and French, 1992)
2. Create the corresponding factor return (e.g., the FF3 value factor)

Step 1 will take place in different places in the code; step 2 will take place in the ap_factors() part of the code

[FernandoRDLL]

I will split it up in like 2 or three issues. I will compare with the FF returns to see how closely they line up

[theisij]

Perfect, thanks! Just to be clear, the FF3 model was just an example. We have characteristics from something like 6-7 asset pricing models (FF3, FF5, HXZ4, HXZ5, SY4, etc.)

[theisij]

You could actually split it into one issue per asset-pricing model. May make it easier to get an overview, and to close and merge issues quickly.