NYC 311 Data Maps the Complaints That Set Neighborhoods Apart
New York City’s 311 records can reveal more than the problems residents share. Aggregated carefully, they offer a portrait of what makes one neighborhood’s complaints stand out from the citywide pattern. A recent analysis applies that idea to 13.5 million reports filed from 2022 through 2025, producing a map that pairs broad complaint themes with the most distinctive specific issue in each residential neighborhood.
The map is a useful demonstration of how public service data can support local analysis, but it comes with a clear caveat: it describes what people report, not everything that happens. Read it as a record of neighborhood attention and reporting behavior as much as a catalog of conditions on the ground.
Turning 311 records into an analytical dataset
New York City’s 311 phone line and website handle requests for city services that aren’t emergencies. The city publishes anonymized records through NYC OpenData. The analysis draws on records from 2010 to 2025, packaged for R as a front end to a large Parquet file managed by DuckDB.
That setup matters because the dataset contains roughly 43 million service requests. DuckDB is designed for analytical queries over large datasets, making it a practical fit for asking questions across years of records. The author works in R, using dplyr-style queries that DuckDB can run against the data. Each record includes fields such as the request date, agency, complaint category, location, borough, and status.
Scale doesn’t make the data self-explanatory. Complaint categories changed over time, and some were substantially reworked. A change in the city’s data management provider also affected measurement. Reporting habits matter too: some residents submit many requests, and the web portal made it easier to file a complaint than calling 311. Those factors can distort comparisons if they’re ignored.
Comparing neighborhoods on a consistent basis
The analysis groups requests by New York City’s Neighborhood Tabulation Areas, or NTAs. These areas sit between census tracts and the city’s 59 community districts in size. The city has 262 NTAs, of which 197 are residential. The author links to the city’s Neighborhood Tabulation Areas data and a reference on the census tract unit, as well as background on New York City community districts.
To reduce the influence of frequent reporters, the analysis counts distinct addresses within each complaint category and residential NTA, rather than simply tallying every submitted request. It also focuses on 2022–2025, after harmonizing complaint categories across the larger historical dataset. The period contributes about 13.5 million records. In the resulting counts, each address is counted once per category it reported, yielding roughly 4.6 million address-category pairs from about one million unique addresses.
The address records don’t include identifying details such as house or apartment numbers. On average, an address appears in about four or five complaint categories across the four-year window. This method reduces the weight of residents who report the same type of issue repeatedly, while still allowing one address to register different kinds of concerns.
Finding what is unusually common
Many complaints are common throughout the city: heat and hot water outages, noise, and illegal parking are examples. A neighborhood’s largest raw count may therefore say little about what distinguishes it from other areas. The analysis instead compares each category’s share of complaints within an NTA with that category’s share citywide.
The category that most exceeds its citywide share becomes the neighborhood’s signature. The process runs at two levels. At the broad level, categories include Sanitation and Dumping, Parking, Driveways, and Vehicles, and Buildings and Construction. At the most detailed level, examples include Traffic: Drag Racing, Appliance: Refrigerator, and Lost Property in Taxi or For-Hire Vehicle: Wallet.
The resulting visualization uses color to show each area’s leading broad category and labels it with the most distinctive specific complaint. That combination gives the map a quick overview while retaining the unusual details that make individual areas memorable.
A neighborhood portrait with reporting bias built in
The labels invite a playful reading of the city. The author points to a nightlife corridor from the Upper East Side to Lower Manhattan associated with reports of lost wallets or keys, brownstone Brooklyn’s bicycle complaints, and South Brooklyn reports involving curb cuts and illegal carports. Woodside, Elmhurst, and Corona show up in connection with public drinking or abandoned vehicles. Forest Hills stands out for reports classified as “Other Non-Emergency Police Matters.”
Those patterns are vivid, but the map’s measurement choices set limits on what they mean. A high relative share doesn’t necessarily mean that a neighborhood has more of a problem in absolute terms. It means that the category accounts for a larger portion of that neighborhood’s address-level complaints than it does citywide. And because the unit is a reported complaint tied to an address, the result reflects residents’ willingness and ability to report, alongside the underlying issue.
The long-term dataset adds another layer of caution. Category definitions and data management changed, which is why the analysis narrows its comparison to four recent years and harmonizes categories. The web portal also made submitting requests easier, while frequent reporters could otherwise dominate simple totals. These decisions make the neighborhood comparison more interpretable, though they can’t remove every source of uneven reporting.
As a piece of urban data analysis, the approach is more revealing than a ranking of complaint totals. It asks what is distinctive after accounting for the citywide baseline and dampening repeat reports from the same address. The resulting map can spark questions about local patterns, but it shouldn’t be treated as a definitive scorecard of neighborhood life. Its strongest use is as a starting point: a way to see where residents’ reported concerns diverge, then investigate what the records can’t explain.
The original analysis appeared on R on kieranhealy.org and was contributed to R-bloggers. The author’s post invites readers to comment on the original blog entry. R-bloggers offers daily e-mail updates about R, learning R, and related topics. The site also links to R and data science jobs, and provides ways to share a blog or submit content through R-posts. Readers can report a page issue here.
===