The rail portion of this ballet, again in theory, should take about 8 minutes. The last trains out of Park Street are scheduled out between 12:45 and 12:53 (the later times because four Green Line trains have to all leave in succession on a single track). The system can then be shut for the night, leaving a bit more than three hours for track maintenance before the first trains the next morning.

Unfortunately, in practice, that’s not how it works. As Marc Ebuña tweet-stormed recently, it takes a whole lot longer. And this costs the T a lot of money.

The last train connection is not guaranteed for passengers to Heath Street (who can take the 39 bus, which is held for connections at Back Bay), but it is guaranteed for Lechmere. Since there’s no layover at Heath Street (since the Arborway terminus was abandoned), these trains have to turn back in to layover at Lechmere. The last train to Heath Street leaves Park at 12:30, arrives Heath at 12:47, and turns back to Park, with a scheduled arrival of 1:06. (In 2007, this train was scheduled 10 minutes earlier, and the “w” note was not present as recently as 2013, although I believe the T has guaranteed these schedules before then.) And this train is given the “w” notation, so that while every other train should be—if they’re on schedule—ready to depart at 12:45, they wait for another 21 minutes before making this connection. And if the Lechmere train is late? The trains still wait. On September 4, for example, the Red Line waited 40 minutes.

This letter costs the T at least $3 million per year. The “w” notation reads: “Last trips wait at some stations, primarily downtown, for connecting service. Departure times are approximate.”

This is entirely unnecessary. Earlier Lechmere trains are just as able to make the connections. There’s a train scheduled to arrive Park at 12:41. If this train were the “w” train, it could drop passengers at Park and continue to Lechmere; any later train could pick up any passenger waiting, but connections for arriving passengers on such a late train would not be guaranteed, other than for buses meeting this train at Lechmere. Or these trains could be operated as non-revenue services, and the 39 bus, which connects inbound with trains at Copley, could provide this service. Other than a few late riders inbound on the E Line, no passengers would be adversely impacted, while every other passenger on the system waiting at least 20 extra minutes (the earliest the last Red Line has left Park Street in the past 30 days has been 1:09, the average has been 1:19—thanks for the coding from @MBTAinfo) would benefit. I’m usually not one to advocate for earlier service, but in this case, either publish a later, more truthful schedule, or run the service on time.

Then there are the costs, which cascade very quickly through the system, since the single Lechmere trip which operates late causes trains on every other line, and 56 bus trips, to all experience delays of at least 20 minutes. But the operators still get paid (overtime, in fact) and the power stays on and the inspectors keep the stations open and the operations staff stays on duty until the last trains pull in. It costs nearly as much money to keep a train stationary as it does to run it, and with overtime, it may cost more. In 2014, the T reported that a bus cost $178 per hour to operate, a heavy rail car $240 and a light rail car $264. Giving them the benefit of the doubt that they’re operating single car green line trains, the cost per hour of four light rail cars (B, C, D and Mattapan), 30 heavy rail cars (five trains at six cars each) and 56 buses comes to $18,224 per hour, or $304 per minute. If the Lechmere car causes a 21 minute delay (as scheduled), this costs $6,384 per day, or $2.33 million per year. In fact, the average delay is more on the order of 34 minutes, which costs $3.77 million. This assumes that all delays are caused by the late Lechmere car; if we attribute 20% of the delays to other causes, there is still a direct operating cost of 1.9 to 3 million dollars per year. (These costs are likely even higher now.)

Two to three million dollars. All because of a “w” on the schedule.

But it gets worse. The T has precious little time between the end of service and the first trains the next morning; most lines aren’t scheduled to be cleared for powering down or work until about 1:30, leaving only about three hours and thirty minutes until service starts in the morning. A 20 to 30 minute delay accounts for 10 to 15% of this time, meaning work crews have to wait for this unnecessary delay before performing maintenance.

Then there are the passengers. If you take the last train, the schedule—and any real time data—will show it coming at a certain time. But you’ll either wind up standing on a platform for 20 to 30 extra minutes, or sitting on the train downtown for that amount of time. There is probably significant ridership loss from people who know how long the wait takes, and choose another mode. This fare revenue is probably minimal in relation to the operating costs, but certainly not zero, but the impact to passengers is more drastic. If we assume just 10 passengers per rail car and two additional passengers on each bus who don’t transfer from a train), it amounts to 450 passengers each inconvenienced by 20 to 30 minutes. That adds up to 150 to 225 hours per night, or 50 to 75 thousand hours per year.

The FMCB’s response to a budget gap has been to push privatization, which is not guaranteed to fill any such gap, but will draw the ire of the unions and potentially degrade service. Yet various measures which this page has noted have fallen upon deaf ears. There’s a lot of money to be saved at the T. There’s a lot of very low-hanging fruit. (Like publishing a set of schedules without a “w” for the 12:47 departure from Heath Street. And who reads paper schedules, anyway?) This problem would be very easy to fix: the next schedules would be amended with a different note for the E Line, perhaps “x: last trip making connection downtown departs Heath Street at 12:25.”


Making that change would go a long way towards paying for real, actual overnight service.

[Thanks to James Jay for noticing this, Marc Ebuña for burning the midnight oil, @MBTAinfo for the code and Stefan! for the maps.)