Critical Inaccuracies in Public Advocate Jumaane D. Williams' Landlord Watchlist

The Landlord Watchlist, hosted on Public Advocate Jumaane D. Williams' official website, is intended to hold landlords accountable for Housing Maintenance Code violations. Recent analysis, however, reveals significant inaccuracies that misrepresent landlord records. A key issue is the case sensitivity of the underlying HPD dataset, which leads to skewed statistics and compromises the credibility of the Watchlist.

One glaring example is the record for Margaret Brunn , president of A&E Real Estate. Although the Watchlist lists her as the 18th worst landlord with 671 open HPD violations between the months of November 2023 and October 2024, a closer examination of the data – adjusted to account for case sensitivity errors – tells a very different story. Margaret Brunn's records report 9,897 open HPD violations between November 2023 and October 2024. Brunn currently has 20,211 open violations total.

In contrast, the Public Advocate’s Watchlist ranks Barry Singer as the #1 worst landlord with 1,804 open HPD violations between the months of November 2023 and October 2024.

The root of these errors lies in the inconsistent formatting of names. Variations such as “MARGARET BRUNN” written in different cases lead to fragmented records and vastly different counts. Without incorporating fuzzy matching or standardizing the data input formats, the dataset cannot reliably consolidate records. This issue could be resolved with a relatively simple change in the code that would standardize all entries, ensuring that variations like capitalization do not result in misreported data.

When analysts aggregate data by name in order to determine a total count of violations or rank landlords, there is an expectation that the name won’t need to be queried in every casing combination possible. Both “Shane Rajcooar” and “Nicholas Quinn” can each be written in 8,192 different ways when you consider every possible combination of uppercase and lowercase letters — plenty of room to hide!

I have raised these issues with the NYC Open Data Help Desk as well as HPD on numerous occasions. In Inquiry #11646, the department acknowledged the problem and indicated that they would review the necessary technological changes to improve data usability.

The fields for “firstname” and “lastname” are treated as case sensitive. If someone is entered as “John Doe” in one record and “john doe” in another, the system treats these as two completely separate individuals. This fragmentation makes it impossible to aggregate or analyze data accurately, since minor differences in capitalization result in multiple, disjointed records for the same person. I would like to also emphasize that the registration contact ID (registrationcontactid) is assigned on a per-building basis rather than being unique to the individual. An individual acting as a registered agent at multiple properties ends up with multiple distinct IDs. This prevents consolidating their actions (like tracking all violations across buildings) and introduces a significant margin of error.

This makes it incredibly easy for corporate landlords with an outrageously high number of violations to go undetected when being listed on something like the Public Advocate’s Landlord Watchlist, and they do.

Above (Fig. 2A) is an example of capitalization currently present in these crucial datasets. Together with case sensitive querying, the ability to consolidate public records accurately is eliminated.

Because of these issues, the same registered agent might appear in the dataset multiple times under different variations of their name (John Doe, John doe, JOHN DOE, john doe) and the discrepancy has been taken advantage of by landlords on countless occasions (see our directory). Given the fact platforms built using this dataset standardize their queries (First Last, and not FIRST LAST), developers & watchdogs using this data for analysis or to produce accountability measures are left with an almost unusable dataset.

In plain language, this means that without implementing technologies like case insensitivity and fuzzy matching (which would standardize the input so that “JOHN DOE” and “john doe” are recognized as the same entity), any analysis, accountability measures, or reporting based on this dataset is fundamentally flawed. Moreover, the fix is easy enough to implement, and yet a tax-funded agency drags their feet at the expense of New Yorkers. It is so easy, in fact, that we decided to host these datasets on our own servers and implement the corrective measures ourselves (see our API documentation) — we are not funded by taxes.

These discrepancies in our city's datasets not only compromise the integrity of vital data, but also hinder meaningful accountability. These datasets are widely implemented across varying systems. By standardizing data entry and implementing fuzzy matching, these inaccuracies can be remedied, ensuring a fairer and more accurate representation of landlord performance. It is imperative that HPD as well as Public Advocate Jumaane D. Williams address these issues promptly to maintain public trust and effectively hold landlords accountable.