A national dataset of property transactions (assessor and deed records) merged with HMDA (Home Mortgage Disclosure Act) homebuyer data. The R code and technical documentation below create the dataset, which lets researchers observe the income, race and ethnicity of homebuyers moving into specific properties and neighborhoods.
Updated code (2026) - hmda_property_merge_public_2026.zip below is a documented, modular R pipeline covering HMDA 1995-2023. It downloads the public HMDA data (CFPB Data Browser for 2018-2023, historic files for 2007-2017, National Archives for 1995-2006), standardizes any standard property dataset (ATTOM, ZTRAX, CoreLogic or county assessor files) through a simple column mapping, assigns census tracts by vintage, and runs a seven-round matching cascade on loan amount, census tract, property value and exact/fuzzy lender name. The zip includes a README with full replication instructions.
Legacy code (1995-2016, ZTRAX) - full_merge_11062018.r and hmda_merging_stats.pdf are the original merge script and technical summary documenting the methodology (following Bayer, McMillan, Murphy & Timmins, 2016) and match rates of 48-57% of mortgage-financed transactions.