Replication package: The Shape of the Tax — Statutory Incidence, Notches, and the Design of New York City's Mortgage Recording and Transfer Taxes

Replication package for The Shape of the Tax: Statutory Incidence, Notches, and the Design of New York City's Mortgage Recording and Transfer Taxes, by Pablo Loschi, independent researcher, Berlin. Version 1.3 corrects three errors that an independent audit found in version 1.1 on 23 September 2026. An intermediate version 1.2 was prepared but never released, and is superseded. Costing geography. The revenue-neutral replacement (Table 6) now uses the four boroughs the paper describes. Staten Island purchases, all recorded without financing because ACRIS does not cover Staten Island, had enlarged the base. The sample goes from 290,840 to 249,395 purchases, and the replacement rates move from 0.81 / 1.50 / 1.27% to 0.88 / 1.57 / 1.32%. The $246.1 million annual carve-out is unchanged. The $500,000 threshold. New York City's transfer tax charges its lower rate on consideration of $500,000 or less, so a price one dollar below $500,000 avoids nothing. The threshold's charm share therefore measures round-number pricing, not avoidance, and it is no longer used for inference. Every state threshold from $1 million to $25 million applies at the threshold itself and is unaffected. The equality claim. The claim "who pays matters, how much does not" is withdrawn. A response proportional to the buyer's rate is rejected, but equal responses are not established, and the design does not settle whether statutory incidence matters. Also in 1.3. Two verification scripts that could not run from the deposit have been fixed, or moved to the provenance folder. The text-consistency check now runs 38 checks, including guards against the withdrawn wording; the arithmetic audit runs 88 checks. Outputs other than the costing are unchanged. Contents. Code and data that reproduce, from the included data, the statutory recording-tax burden (Table 1), the 2019 notch experiment (Tables 2–5, Figure 2), the loan-to-value test, and the costed replacement (Table 6). The record-linkage scripts are included for provenance.