One row per PACT development with project name, status (Construction Complete, Under Construction, Planning and Engagement), conversion date, total units, developers, general contractor, and property manager. Used as the master list of active PACT developments and the source of unit counts and manager attribution.

Building-level records for NYCHA-managed properties including address, borough-block-lot (BBL), latitude, longitude, and a privately_managed flag. Used initially for BBL lookup, but PACT buildings are removed from this dataset after transfer to private management. Superseded by PLUTO for the control group BBL set. Primary source for the NYPD analysis non-PACT spatial index (Phase 0b): all ~2,955 building-level coordinates are fetched to generate per-building 150m query circles for the non-PACT incident count.

One row per NYCHA development with unit counts, borough, cross-street locations, and critically, rad_transferred_date — the date each development was transferred under RAD/PACT. This date is the basis for each year's changing PACT/non-PACT denominator. Also used for geocoding development locations via cross-street intersections.

Tax lot–level property data for all of NYC with owner name, residential unit count, address, and coordinates. Used in three ways: (1) resolving BBLs for PACT developments by searching for the PACT entity's owner name (e.g., "NYC PACT PRESERVATION PARTNERS LLC"); (2) identifying all non-PACT NYCHA buildings as lots where ownername contains "NYC HOUSING AUTHORITY" — the PLUTO owner-name split cleanly separates PACT from non-PACT because ownership transfers at conversion; (3) providing per-BBL lat/lon coordinates for the NYPD analysis PACT spatial index (Phase 0a), where each BBL is queried by BBL number to retrieve building coordinates used to center the 150m incident query circles.

Free geocoding API that converts address strings and intersections to coordinates. Used as a fallback to geocode cross-street intersections from S3 for developments where PLUTO owner-name search failed, enabling spatial bounding-box queries on PLUTO.

One record per marshal-executed eviction, with BBL, marshal docket number, court index number, address, execution date, and residential/commercial indicator. This is the primary numerator for the eviction rate calculation — filtered to residential units from January 2022 onward. The BBL field enables direct joins to PLUTO-derived PACT and non-PACT lot sets. The docket number enables a join to OCA court case records (S7).

Pre-parsed CSV tables derived from the raw OCA XML feed received via SFTP from the NYS court system. Three tables used here: oca_warrants (contains the marshal docket number as enforcementofficerdocketnumber, enabling the join to S6); oca_index (case-level classification: Non-Payment, Holdover, and disposition method); oca_causes (cause of action type). Joined to marshal data via docket number to add case type to each execution. Note: the HDC tables contain respondent mailing ZIP codes but no street-level property addresses, so building-level filing rate analysis is not possible through this pipeline.

One record per DOHMH proactive rodent inspection, with BBL, inspection date, inspection type (Initial, Compliance, etc.), and result (Passed, Failed for Rat Activity, Bait Applied, etc.). Used to compute annual rat activity failure rates at all PACT developments (Construction Complete and Under Construction) vs. non-PACT NYCHA sites. Only inspections at a PACT development on or after its rad_transferred_date are attributed to the PACT group; pre-conversion inspections are excluded. Note: these are proactive city inspections, not tenant complaints — inspection frequency varies by site and program enrollment, which can affect apparent failure rates independent of actual rodent conditions.

One record per 311 service request city-wide, with agency, complaint type, BBL, address, and created/closed dates. Filtered to agency = 'HPD' (Housing Preservation and Development) for housing-specific complaints such as Heat/Hot Water, Plumbing, Paint/Plaster, and Elevator. Used exclusively for PACT developments because non-PACT NYCHA tenants route complaints through NYCHA's own system (MyNYCHA), making cross-group 311 comparison invalid. Data covers 2020–present; only post-conversion complaints are attributed to each PACT development.

One record per DOB job application, with BBL, job type (New Building, Alteration Type 1/2/3, Demolition), filing date, and initial_cost — the owner-declared construction cost estimate. Used as a capital investment signal at PACT developments: alteration job filings post-conversion indicate renovation and capital improvement activity. Limitation: NYCHA historically filed fewer DOB permits than private owners due to sovereign immunity claims, making direct PACT vs. non-PACT comparison difficult.

HUD's Real Estate Assessment Center (REAC) scores public housing authorities annually on a 0–100 physical conditions scale. Two separate systems apply here: PHAS (Public Housing Assessment System, retired 2023) scored NYCHA as a single PHA (NY005) on a composite of physical, financial, management, and capital fund components — producing one score for the entire authority per inspection cycle, not per development. NSPIRE (National Standards for the Physical Inspection of Real Estate, introduced 2024) replaced PHAS with a more stringent physical-only protocol; scores are not directly comparable across the system change. For PACT-converted properties, REAC Multifamily inspection scores are published in a separate Excel file (MF-Inspection-Report.xls) matched by property name — no BBL or NYC address field is available.

Why this data is not charted here: Investigated as a potential physical conditions metric, but three problems make it unsuitable for this analysis. (1) Sparseness: NYCHA has only five years of recorded PHAS/NSPIRE scores between 2015 and 2024 (two COVID-waiver years, a system transition in 2023–24). (2) Incomparability: PHAS includes non-physical components (financial, management, capital fund) and NSPIRE uses a stricter physical protocol — the two cannot be placed on the same axis. (3) PACT match quality: the multifamily REAC file has no BBL or address; matches to NYCHA development names are approximate and could not be confirmed. Only three Brooklyn developments were tentatively matched. See methodology for full documentation of this investigation.

Two Socrata datasets cover all NYPD complaint incidents (felony, misdemeanor, violation) reported across NYC. The Historical dataset (qgea-i56i) covers 2006 through the prior calendar year; the Current YTD dataset (5uac-w243) covers the current calendar year. The geospatial field name differs between them (lat_lon vs. geocoded_column), affecting how within_circle() SoQL queries are structured. Analysis begins in 2015.

Both PACT and non-PACT groups use a uniform 150m-per-building circle methodology. For each building, a within_circle(field, lat, lon, 150) SoQL query is issued against both datasets; results are deduplicated by cmplnt_num. No prem_typ_desc filter is applied — the geometry defines scope. Scattered developments (LINDEN: 15 parcels; BUSHWICK II: 9 parcels) use one circle per BBL rather than a centroid to avoid over-sweeping. Non-PACT building coordinates come from S2 (~2,955 buildings); PACT coordinates from S4 (PLUTO, by BBL).

PACT data covers 5 years pre-conversion and all post-conversion years, producing two charts: (1) a cohort calendar-year chart that tracks incident rates for PACT-destined developments (using pre-conversion data before transfer, post-conversion data after) alongside non-PACT rates from 2015 to present; (2) a before/after bar chart using anniversary-year buckets anchored to each development's conversion date (bucket 0 = first 365 days post-conversion, bucket −1 = 365 days before, etc.), so each bar represents exactly one year of exposure. Rates are per 1,000 residential units with Poisson 95% confidence intervals.