What New York City eats, by neighborhood

×

How the city defines a cuisine

Short version: it doesn't, really. The cuisine label is not assigned by a city analyst against a rulebook — it is chosen by the restaurant.

What the city's own data dictionary says

The Health Department defines the field in exactly one sentence:

“This field describes the entity (restaurant) cuisine. Optional field provided by restaurant owner/manager.”

So the value is self-reported and optional. When a restaurant is permitted, its owner or manager picks one label from a fixed list of roughly 90 choices the city maintains. There is no published definition of what separates one category from another, and no review of whether the pick is “right.”

One label each — and only one. We checked: none of the 31,217 establishments carry more than one cuisine label. So the categories never overlap and the percentages never double-count. But it also means the single list is forced to mix two different ideas — a country of origin (Chinese, Italian) and a kind of food (Coffee/Tea, Pizza, American) — and a place can only be one. An American-style coffee shop is filed as “Coffee/Tea,” never also as “American,” so “American” undercounts American-style food. Use the place of origin / food type filter on the charts to look at one dimension at a time.

Why that matters here

• The categories are uneven by design. Some name a country (Chinese, Mexican), some a region (Caribbean, Eastern European), some just a food (Pizza, Donuts, Sandwiches), some a style (New American, Fusion). They sit in one flat list with no hierarchy.
• Because it is optional, about 3,500 establishments leave it blank — they are real restaurants with no cuisine label at all.
• A few labels are easy to misread. “Chinese” covers every regional Chinese cuisine; “Spanish” in this dataset means Spanish-Caribbean/Latino food, not food from Spain.

That is why this page treats the labels as the city's own coarse, self-reported tags — useful at scale, but not an official taxonomy. Source: NYC Open Data dataset 43nn-pn8j, column data dictionary.

×

Methodology & sources

Data source

New York City Department of Health and Mental Hygiene, DOHMH New York City Restaurant Inspection Results (NYC Open Data, dataset 43nn-pn8j).

How the city defines “cuisine”

Per the dataset's data dictionary, the cuisine field is “an optional field provided by the restaurant owner/manager” — self-reported, picked from a fixed list of about 90 values the city maintains. The Health Department publishes no written definition distinguishing one category from another, which is why the labels are uneven and roughly 3,500 establishments are left blank.

Counting method

• The dataset has one row per inspection violation. Each establishment carries a unique CAMIS id. We count distinct CAMIS, so each restaurant is counted once regardless of how many inspections or violations it has.
• Each establishment is assigned to a borough and a 2010 Neighborhood Tabulation Area as recorded in the dataset.
• “Share of neighborhood” = a cuisine's establishments divided by all establishments in that neighborhood (any cuisine, including blank/unlabeled).

Neighborhood metrics

• Most mixed uses the Simpson diversity index over a neighborhood's labeled restaurants: 1 − Σ(share of each cuisine)², i.e. the probability that two restaurants picked at random are different cuisines. Computed only for areas with at least 50 restaurants.
• Most dominated is the single largest cuisine as a share of all the area's restaurants (same 50-restaurant floor).
• Per resident divides restaurant count by 2010 census population (the matching geography year); non-residential areas like parks and airports are excluded. Population is from 2010, so this is an approximation.
• Local specialty is the cuisine an area over-indexes on most versus the city — a location quotient: (cuisine's share of the area) ÷ (cuisine's share of the city). Only cuisines with 5+ local spots and a quotient of at least 1.3 qualify; areas with no standout are left blank.
• Chains are identified by business name: any name with 5+ locations citywide (after normalising spelling, so “DUNKIN” and “DUNKIN'” merge). Caveat: this also catches independent shops that happen to share a generic name — most notably “Kennedy Fried Chicken,” which is not a real franchise. So chain shares are slightly overstated in areas with many such shops.

Geography

2010 Neighborhood Tabulation Areas, New York City Department of City Planning. Boundaries simplified for display. Population: 2010 U.S. Census via NYC Open Data dataset swpk-hqdp.

Known limitations

• Cuisine is a single coarse, self/inspector-assigned label.
• About 3,500 establishments have a blank cuisine field; roughly 600 are filed as “Other.”
• Recently-closed establishments may still appear.
• Neighborhood boundaries do not always match colloquial enclaves.
• The restaurant dots cover the roughly 97% of establishments with usable coordinates; about 800 (2.6%) lack a geocode and are not shown. These skew toward concessions inside airports, stadiums and food courts rather than food carts — mobile vendors are inspected under a separate program and largely absent from this dataset.
• Locations are the city's own geocoding, and a share of them are wrong. When a street name exists in several boroughs (“5 Avenue,” “Broadway,” etc.), an address can resolve to the wrong borough. A known example: New China Tung, a Brooklyn restaurant, is recorded by the city at 140 5th Avenue in Manhattan — its 718 phone number gives away the true borough. The map plots whatever coordinates the city provides, so it reproduces these errors and a few restaurants land in the wrong neighborhood. As a rough gauge, about 1,450 of 12,500 Manhattan-listed establishments carry outer-borough phone numbers (a noisy signal, not a precise error count).