Methodology: US Home Address Privacy Report 2026

Dataset A is a 50-state review of address confidentiality and privacy protection rules and programmes. It is structurally complete for all 50 states, but every state row remained NEEDS REVIEW for publication clearance. It may therefore be used descriptively only, and no VERIFIED-only state prevalence figure is derived from it.

Dataset B is the audit of local public-record systems. It covers 50 local jurisdictions with four mandatory record categories each - property assessor/property records, property tax, recorder/land records and court records - giving 200 local audit rows in total. The four categories cover the record types most likely to carry a home address.

Dataset C is a 20-jurisdiction deep dive into how state-level rules are implemented at local level, scored with the 8-point Privacy Guidance Clarity Score described below.

The Dataset B sample was selected purposively for geographic spread and intended size distribution. It was not hash-selected and it is not a probability sample. Accordingly the report publishes no margin of error, no confidence interval and no estimate of national prevalence. Every percentage describes the audited rows only, and Dataset B is not statistically representative of U.S. jurisdictions.

Dataset C selection, and only Dataset C selection, used an outcome-blind procedure applied to the frozen Dataset B pool. Each candidate jurisdiction was given a fixed identifier, a SHA-256 hash of that identifier was used to order candidates within pre-set region and intended-size quota groups, and the quotas and ordering were fixed before the deep-dive research began. Because the hash ordering is unrelated to what any jurisdiction publishes, selection could not be nudged towards jurisdictions with more convenient results.

Where several audited jurisdictions rely on the same platform or the same published policy, those shared systems and policies were recorded with identifiers and deduplicated in the analysis, so a single statewide system is not counted as several independent findings.

The study used an official-source-first standard. Coding had to rest on the official page, form, statute or documented procedure published by the responsible body. Search-engine snippets were treated as discovery only and were never used as evidence for a coded value.

• No person searches were performed.
• No accounts were created.
• No payments were made.
• No access control, paywall or restriction was circumvented.
• Where a required source was inaccessible under those rules, the value was coded Unknown.

Each row carries one of three statuses, which were preserved rather than merged:

• VERIFIED - coding is supported by an official source that states the point directly.
• NEEDS REVIEW - a source exists but is ambiguous, partially applicable or internally inconsistent.
• UNVERIFIED - no qualifying official source was located within the protocol rules.

Published percentages use VERIFIED rows with known coding only, which is why the denominators differ between figures.

Unknown means the documentation did not allow a determination. It does not mean the requirement is absent. Unknown values were retained in the dataset and excluded from denominators rather than recoded, because converting Unknown to No would inflate apparently favourable results.

Dataset C scores each jurisdiction out of 8, awarding one point for each documented component:

1. A guidance page covering address privacy or confidentiality.
2. A dedicated process or form.
3. Stated eligibility criteria.
4. A list of required documents.
5. A stated submission method.
6. Stated treatment of existing records.
7. Stated treatment of historical records.
8. Contact information for questions.

Proposed headline findings were reviewed in a separate audit step against the frozen protocol. Findings that did not meet the publication rules, including any figure that depended on recoding Unknown values or on non-verified rows, were withheld or rejected rather than promoted. Exclusions were documented.

The summary coding behind each published figure, including numerators, denominators and verification basis, is available as a CSV download. A PDF edition of the report is also available; the HTML report remains the canonical version. Return to the report.