Research · Evaluation · Updated Oct 7, 2026

Do more trains bring more riders? More trains cut waits by about half a minute, with no clear rise in riders.

In 2023 and 2024 the state gave the MTA $35 million to run trains more often on 12 lines, at middays, in the evening and on weekends. On weekdays those lines ran about 14% more trains, and the average wait fell by about half a minute. Riders at their stations rose 1.2% (range −0.7% to +3.2%): any rise was small, no more than about 3%, a quarter of the rise in trains. On weekends, track work often kept the extra trains from running.

Trains came more often; any rise in riders was small

Trains per hour and riders entering, at stations where the lines given more trains run most of the trains, against stations on lines that got none, by month before and after the extra trains began. Weekday middays and evenings.

New York ran more subway trains, with no clear rise in riders

Trains per hour and riders entering, 103 stations on lines given more trains against 199 others, weekday middays and evenings, extra trains added 2023–24, %, with 95% intervals

Source: MTA subway train feeds; MTA subway hourly ridership

Each month is measured against the six months before the extra trains began, so those months sit near zero by design.

Waits fell by about half a minute

Average wait for a rider who turns up at random, on each line given more trains, in the six months before and after. The wait counts uneven gaps: when trains bunch, more riders arrive during the long gap.

Extra New York subway trains cut waits by about half a minute

Average wait for a rider arriving at random, each line given more trains, six months before and after, 2023–24, minutes

    Source: MTA subway train feeds

    Extra trains, as announced and as run

    Each increase the state and the MTA announced, with the trains per hour we measured on the line in that time band, six months before and after. Most extra weekday trains ran as planned. Extra evening trains on six lines, and some weekend ones, did not show up.

    LineWhenFromAnnouncedTrains per hourChangeRan?

    Trains per hour: each direction, at the stations the line serves most days, the median month. Middays are 10 am to 4 pm, evenings 7 to 10 pm, weekends 10 am to 7 pm. “Ran” means the line ran at least 8% more trains. Announced: every so many minutes, from the Governor’s and the MTA’s releases.

    On weekends, track work got in the way

    Share of Saturdays and Sundays in the year after the extra trains began when the line ran at about its new frequency, 10 am to 7 pm.

    Extra weekend subway trains ran under half the time on four of seven New York lines

    Weekend days each line ran at least 90% of its promised trains, 10 am to 7 pm, in the year after more trains began, 2023–25, %

      Source: MTA subway train feeds

      Riders moved as much in hours with no extra trains

      At the same stations, riders rose about as much in the weekday morning rush, which got no extra trains. And moving every change a year early, before any extra trains ran, gives −2.0%, larger than the gain itself. So the small rise in riders could be chance.

      Subway riders moved about as much in the morning rush, which got no extra trains

      Riders entering at 103 New York stations on lines given more trains, against 199 others, by month, extra trains added 2023–24, %, with 95% intervals

      Source: MTA subway hourly ridership

      Riders after the extra trains, every way we checked

        What could change this answer

        Chance alone moves this comparison by about 2%, so the 3% bound is looser than its range suggests. The rest would mostly hide a gain or flatter one.

        • Chance swings of about 2% (either way)Moving every change a year early, before any extra trains ran, gives −2.0% (−3.6 to −0.5). With swings that size, the +1.2% can’t be read as an effect.
        • The 2023 changes alone (would shrink it)Leaving out the 2024 D, J and M trains gives +3.0% (+1.4 to +4.8). That is one of the 13 comparisons above, and with that many, one or two will look clear by chance. The MTA also picked lines and hours where riders were coming back fastest, which would make it too high.
        • One year may be too short (would hide a later gain)Riders may take longer than a year to change habits, such as where they live or work.
        • Entries, not trips by line (either way)Entries count people who pay at the turnstile, not trips, and fare evasion changed over these years. Extra trains on one line may draw riders from another line at the same station, which this can’t see.

        What it means

        • What the data showThe $35 million bought shorter waits, about half a minute on weekdays, for people already riding. Within a year it brought no clear rise in riders: no more than about 3%, against 12% more trains at their stations. On weekends, track work often kept the extra trains from running.
        • What follows, if it holdsMore trains alone may not win riders back within a year; habits may take longer to change. On this evidence, the case for more off-peak trains rests on the time they save people already riding, not on new riders. The data can’t say whether that time is worth $35 million.

        How this works

        Each increase started on a known day, on some lines and in some hours only. Stations on those lines are compared with stations on lines that got no extra trains in the same hours, over the same months.

        • TrainsEvery train leaving every station, from the MTA’s live train feeds, kept since April 2021 (our open data). For each line, hour and day: the trains that left and the gaps between them. Days when the feed itself broke, holidays and the week around Christmas are left out.
        • WaitsFor a rider who turns up at random, the average wait is the sum of squared gaps over twice the sum of gaps. With even gaps that is half the gap; bunching makes it longer. Waits are for that line alone, averaged over its stations.
        • RidersThe MTA’s hourly entries by station, from fare taps and swipes, averaged over the days of each month in the hours given more trains. The main comparison takes stations where those lines run at least half the trains in those hours, and weights stations by their riders.
        • The comparisonEach station’s riders against its own average in the six months before, minus the same for stations whose lines got no extra trains in that time band within a year either side, matched by borough (and Manhattan below 60 St). Months 0 to 11 after the change count. Months when one of those lines was cut by work to under 85% of its usual trains are left out, and the G’s window ends before its summer 2024 shutdown.
        • Riders per trainThe change in riders over the change in trains per hour at the station, in percent. 0.2 would mean 10% more trains brought 2% more riders.
        • Intervals95% intervals from resampling stations 1,000 times.
        • How sure we areWe grade each answer high, medium or low. High: the range is narrow and every check agrees. Medium: the answer holds, but its size depends on a choice we made, or a check disagrees. Low: the answer rests mainly on an assumption we can’t test.

        Stations where a line stopped calling for a month or more in the window (the M left Queens Blvd from September 2023 into 2024) are left out. Extra weekend trains are left out of the headline: weekend riders swing too much from month to month to read. The 2024 changes sit close to congestion pricing (January 2025), though the result is the same without the Manhattan core. The caveats that matter most are under What could change this answer.

        What to watch next

        A second year of riders after the 2023 changes would show whether riders follow more slowly than a year allows. The MTA’s estimates of trips between stations, from 2024 on, could show whether the extra D, J and M trains moved riders from other lines. Every number here is in subway-service.json, so you can check our work.