NYC Housing Data · Independent research
Explore recorded transactions across the five boroughs. See how deed amounts, mortgage matches and tax rules shape the numbers people use to understand the housing market.
Public records, 2016–2025. Frozen research data, with sources and limitations.
From Housing Stories
Under New York's 2019 rules, that last dollar adds about $5,000 to the buyer's tax. The sales records show a shift around the line.
Read the story and get the charts
The papers
The amount on the deed, the financing label attached to a sale, and the tax the law assigns to it.
When does recorded consideration follow a legal rule rather than a negotiated price, and what does that do to a price index?
16,655deeds whose recorded amount can follow a legal rule rather than a negotiated price, 2003–2025
Does a recorded cash–mortgage price gap reflect financing, or who buys with cash?
9.25 → 4.87log-point house gap, before and after excluding same-surname deeds
Submitted to Journal of Housing Economics
How did sales shift at the transfer-tax notches created in 2019, and what does that mean for taxing borrowers and cash buyers differently?
7.8 → 1.6sales at exactly $2 million relative to nearby price density, before and after the 2019 reform
Submitted to American Economic Journal: Economic Policy
How Measurement Choices Rank Housing-Access Instruments
If a buyer's certainty of closing, usually a cash buyer's, can decide who gets a home, how would you tell which policy widens access at the lowest cost? The paper compares acquisition caps, fair-share quotas, constant charges and contention-triggered duties in a simulated market, and finds that the answer depends on how access and efficiency are measured.
An efficiency benchmark computed over the buyers still waiting moves with the policy and can reverse the sign of its effect; a two-period example shows why. Access counted over the whole run understates a cohort's gains by up to 2.5 times and reorders the instruments.
The comparison is conditional. Run at New York City proxy values, removing the cash discount lowers the share of owner-occupiers who leave without buying only from 9.8% to 9.5%. The model leaves out strategic bidding, renter welfare and administrative costs. The empirical papers motivate the model without confirming its premise: Who Gets the House? finds that the measured contrast changes with buyer identity and transaction screens; it does not identify the value of closing certainty.
Related methods paper: Auditing policy dependent benchmarks in allocation simulations (PDF) · Code and archive.
Why it matters
Some registry amounts are not prices. Deeds that record a lender taking title pass the usual screens and shift the timing of a repeat-sales index through the 2008 crisis. Classifying deeds by their parties removes them.
Excluding deeds with shared buyer and seller surnames roughly halves the measured house-price contrast. Those names are a proxy for possible related transfers. The result does not identify the effect of paying cash.
Recorded prices shift around new tax thresholds. Observed loans also allow a calculation of gross mortgage charges under the statutory schedule. Actual tax payments and who ultimately bears the cost remain unmeasured.
The data
"Same data" means the papers draw on the same free New York City records. Each paper builds its own sample from them and answers its own question.
Two public sources
Linked record by record
Three samples
The Shape of the TaxPrices and loan amounts: 172,725 financed purchases; 431,471 sales around the tax thresholds, 2016–2025. Who Gets the House?Repeat sales of the same property, with the deed's parties: 6,006 house pairs, 2016–2025. When a Deed Is Not a Market SaleEvery deed and its parties: 1.47 million deeds, 2003–2025, and 238,290 repeat-sale pairs.Sources
All sources are free NYC Open Data. Each paper's archive records the exact extracts it used.
Linked dataset
Deposited on its own (doi:10.5281/zenodo.22977532); it reproduces the financing flags of the tax and house papers. The companion simulation uses two aggregate inputs from the empirical papers; it does not estimate outcomes from transaction records.
Coverage
Staten Island deeds are kept by the county clerk, outside ACRIS. Co-operative apartments change hands as shares, not deeds, so deed-based measurement largely misses them.
Listen or watch
Both were generated with Google NotebookLM from the September releases. The paper pages link to the current releases, including subsequent qualifications.
Audio · 23 min · AI-generated
A conversational overview of the three empirical papers, the companion simulation and the linked dataset.
Video · 8 min · AI-generated
Narrated slides on how the studies, the simulation and the 729,047-sale dataset fit together.
Terms
About
Lawyer (UBA) · Executive MBA in progress (Quantic)
Principal Systems Engineer · Independent researcher · Berlin
Pablo Loschi is a Principal Systems Engineer at Verve in Berlin and an independent researcher in housing economics and tax design. After more than fifteen years building large-scale cloud and data platforms, first in Argentina and for the past seven years in Germany, he brings an engineer's habits to empirical work: documented linkage rules, versioned releases and replication code. The papers distinguish agreement between files from verified accuracy. Trained as a lawyer at the University of Buenos Aires, he reads the statutes behind the records, and each of these papers starts from one.
Housing finance is also a practical interest. He founded HolaCasaClara, a free, independent comparator of Argentine mortgage loans that turns banks' fine print into total borrowing costs; its launch was covered by iProfesional and elDiarioAR, and Ámbito Financiero used it as a source.
