Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Investigate using condo sales val geographies as strata grouping var rather than township #21

Closed
wrridgeway opened this issue Jan 30, 2024 · 5 comments
Assignees
Labels
method ML technique or method change

Comments

@wrridgeway
Copy link
Member

Valuations is handcrafting real estate submarkets for condos for sales validation. It might be worth using these geographies rather than township when constructing condo strata.

@wrridgeway
Copy link
Member Author

@ccao-jardine will you let me know once you've got your hands on this data?

@ccao-jardine
Copy link
Member

In case it didn't find its way to you: Res_nbhd_groups.xlsx

It is only for the City tri. Sales val uses these groupings when the sale is in the City tri; otherwise, it uses township for now.

@dfsnow dfsnow added the method ML technique or method change label Apr 15, 2024
@ssaurbier
Copy link

ssaurbier commented Dec 21, 2024

What is the status on this issue? This sounds like a good way to impute bias in the model.

Manual definitions (ie, neighborhoods) probably reflect arbitrary assumptions about property value, which are otherwise contained in other variables - nhbd is likely a confounding factor here. If there is any information contained within nhbd, which surely there is, it can be sufficiently parsed through independent factors (political representation, distance to transit, local school zone rating, other geo params).

One can imagine dozens of cases of bias - eg, a corner of a neighborhood that is gentrifying due to another neighborhood, and all the homes in that corner are assessed arbitrarily lower due to association with a random square on a map. This could be solved with interaction variables and hierarchical structuring, but to what end? I think it would be better to scrap this entirely. I cannot see how it does not induce systematic error.

More importantly, arbitrary/legacy neighborhoods probably embed historical discrimination and inequalities into assessments. This seems like a potential key source for persistent structural discrimination / racism in the county.

Please close or update this issue.

@ssaurbier
Copy link

Did everyone sign off on the redlining bias here? @dfsnow

@dfsnow
Copy link
Member

dfsnow commented Feb 8, 2025

Oh, I missed the original comment. No, this issue is closed because it's no longer relevant. #101 replaced the strata binning with a simpler building means feature.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
method ML technique or method change
Projects
None yet
Development

No branches or pull requests

5 participants