Add ACS, rent and property taxes and 3-year CPS

PolicyEngine / policyengine-us-data

A package to create representative microdata for PolicyEngine US.

1 stars 2 forks source link

Closed PavelMakarchuk closed 1 month ago

PavelMakarchuk commented 2 months ago

Fixes #31

MaxGhenis commented 2 months ago

Have you tried running it?

PavelMakarchuk commented 1 month ago

Could you add acs/README.md linking to the codebook?

Also if easy, could you add rent and property tax as variables?

Added a readme with documentation

We already include GRNTP -> Gross Rent RNTP -> Monthly rent

I don't see variables for property taxes in the dictionary

MaxGhenis commented 1 month ago

PavelMakarchuk commented 1 month ago

[broken image removed by @nikhilwoodruff- weird gh issue?] No idea how i missed that - added

nikhilwoodruff commented 1 month ago

@MaxGhenis a few things (thanks @PavelMakarchuk; taking over the baton for this PR) you might want to take a look at (but not essential or blocking).

I've moved from our implementation of QRFs in survey-enhance to a more lightweight wrapper over quantile-forest which seems faster and better-written than mine. We still have the mean quantile ability. The new implementation wrapper is in utils/qrf.py.
Seems like we shouldn't calibrate real estate taxes from the SOI? But I can add calibration targets for total rent and property taxes, any ideas on the definitive source to use? They match the ACS totals after imputation.

MaxGhenis commented 1 month ago

Correct we should not calibrate real estate taxes from SOI. See #29 for other options.

nikhilwoodruff commented 1 month ago

SNAP is looking really nice with the rent imputation: after applying USDA's 82% take-up rate, we get 103bn vs 107bn in 2023.

nikhilwoodruff commented 1 month ago

Also tied in #67 here.

nikhilwoodruff commented 1 month ago

Getting memory breaches but I'm going to merge since it ran locally.