Who Gets the House? Related-Party Transfers and the Cash–Mortgage Price Gap in New York City
Abstract
Repeat sales can hold a property fixed while comparing transactions with different financing labels. Using New York City deed and mortgage records for 2016–2025, the paper finds a house price contrast of 9.25 log points [7.17, 11.35] between purchases with and without a matched mortgage, after permit, institutional-seller, party-type and ZIP-year controls. The contrast depends on who the cash buyer is. For company cash buyers it is 1.53 log points [−1.79, 5.44], an interval that includes zero. Excluding 358 pairs in which seller and buyer share a surname reduces the house contrast to 4.87 log points [2.69, 6.76], and transfer reports show non-market signs more often among the flagged deeds. The surname screen is a proxy, so its effect does not measure the causal role of family transfers. Condominium contrasts are small under a common design. AI-assisted document checks of a stratified probability sample of 40 endpoints find purchase borrowing behind two unmatched-mortgage labels and leave 18 unresolved; correcting those two labels moves the contrast from 9.25 to 9.23 log points. Random recoding keeps the gap positive, but errors that depend on price growth can reverse it. The design identifies neither a causal financing effect nor the incidence of the mortgage recording tax. Files: the paper and an update archive. To rebuild the full replication package, run reconstruct_v3_7.py on the version 3.5 archive (record 23049938), as APPLY_UPDATE.md explains. See revision_notes.md for earlier changes.
