Skip to content
Docs

Subway ridership by hour

How many people entered the subway at each station complex, each hour, with each fare type, since January 2017. The MTA's own estimate, copied from data.ny.gov.

Rows
224,430,037
Size
533 MB
Covers
13 January 2017 to about a week ago
Updated
Nightly (the MTA adds to it weekly)
Source
MTA, on data.ny.gov (subway hourly ridership: 2017-2019, 2020-2024 and beginning 2025)
Fileshttps://data.transitlab.nyc/v1/subway_hourly_ridership/year=YYYY/month=MM/YYYY-MM.parquetOne file a month, 117 so far. In the database as transitlab.subway_hourly_ridership.

Rows and size from catalog.json, 2026-10-08.

The entries at one station complex in one hour with one fare type. The MTA estimates them from fare payments (MetroCard swipes and OMNY taps). A complex is a group of stations that share a fare area, such as Times Sq-42 St, so a row counts people who came in through any of its gates.

The table joins three MTA datasets (13 January 2017 to 2019, 2020 to 2024, and 2025 on). They have the same columns and the same labels, so there is no break at the year ends. What changes is the mix: OMNY starts on 31 May 2019, and until then every row is MetroCard. The Staten Island Railway and the Roosevelt Island Tram are in it too: filter on transit_mode = 'subway' for the subway alone.

We changed three things from the MTA’s files: the columns are typed (transit_timestamp is a timestamp, counts are integers), each month is its own file, and we left out the georeference column, which repeats latitude and longitude. Values are the MTA’s.

SQL
-- Subway entries by year, and the share paid with OMNY
SELECT year(transit_timestamp) AS year,
sum(ridership) AS entries,
round(100 * sum(ridership) FILTER (payment_method = 'omny') / sum(ridership), 1) AS omny_pct
FROM transitlab.subway_hourly_ridership
WHERE transit_mode = 'subway'
GROUP BY year
ORDER BY year;

Or read one month straight from its file:

SQL
-- The ten busiest station complexes in August 2026
SELECT station_complex, sum(ridership) AS entries
FROM 'https://data.transitlab.nyc/v1/subway_hourly_ridership/year=2026/month=08/2026-08.parquet'
WHERE transit_mode = 'subway'
GROUP BY ALL
ORDER BY entries DESC
LIMIT 10;
transit_timestamptimestamp

The hour, New York local time. The MTA rounds down: a tap at 1:37 pm is 1 pm.

transit_modevarchar

subway, staten_island_railway or tram (the Roosevelt Island Tram).

station_complex_idvarchar

The MTA's ID for the station complex, as text.

station_complexvarchar

Name of the complex with its routes, e.g. Zerega Av (6). A complex can hold several stations, such as Times Sq-42 St.

boroughvarchar

Borough of the complex.

payment_methodvarchar

metrocard or omny.

fare_class_categoryvarchar

Fare type, e.g. OMNY - Full Fare, Metrocard - Unlimited 30-Day, Metrocard - Seniors & Disability.

ridershipbigint

Riders who entered at that hour with that fare type. An estimate from fare payments.

transfersbigint

Of ridership, riders who entered on a free transfer. A subset: already counted in ridership.

latitudedouble

Latitude of the complex.

longitudedouble

Longitude of the complex.

  • Entries, not trips. A rider counts once, when they go through the gate. A change of train inside a complex is not counted, and transfers covers only free transfers from a bus or another system.
  • It is an estimate. The MTA works the counts out from fare payments and says they are within 1% of the figures it gives its board. Treat small counts with care.
  • Each row is one fare type, so add them up for a total. The mix moves over time as MetroCard gives way to OMNY.
  • Times are New York local time, with no zone attached. The hour is the start of the hour: 8 am covers 8:00 to 8:59.
  • The MTA adds to it weekly and revises it. The last days of the newest month are not there yet, and late fares change past months slightly. Each night we check the current and previous month, and every month on Sundays, and rebuild a month only if its totals changed.
  • Terms: NY open data. Credit the Metropolitan Transportation Authority. Source datasets: 2017 to 2019, 2020 to 2024 and beginning 2025.