Why is off-peak Red Line service being cut?

Answer appears to lie in staffing agreements

IT’S NOT SURPRISING that service on the Red Line has suffered in the wake of the June 11 derailment that took out a large chunk of the line’s signal system. The lack of signals precludes running the normal complement of trains at rush hour—14 per hour. Instead, the T is running approximately 9 to 10 trains per hour during these times. This causes crowding, congestion and full trains, but until the signal system is fully operational, the safe operation of service during these times will require continued reduced capacity.

While it is encouraging to hear the MBTA promising better service by August 15, one option for improving service now would be to run more off-peak trains. Outside of rush hour, 9 trains per hour is the normal scheduled frequency on the Red Line. While the agency is technically capable of running this service now (since this is how many trains are being run now at peak periods), it’s not being done. Since June 11, only 6 trains per hour have been run on the Red Line outside of peak hours, affecting tens of thousands of passengers who now have to wait as long as 20 minutes for trains on the Ashmont and Braintree branches.

No explanation for this service disruption has been forthcoming. By examining data for the past two months, however, we conclude that the likely cause of the drop in off-peak service is not technical. It could be fixed by minor changes to labor and scheduling policies – specifically by shifting some staff from peak hours to midday, or increasing off-peak staffing, until the line is back to full service. While this won’t reduce crowding on peak-hour trains, it will make the Red Line a bit less onerous for commuters at other times of day.

Last Wednesday, the T tweeted out that they have made progress on fixing the signal system and have added trains to service. While it is true that rush hour service levels have improved slightly, the capacity of the Red Line still hovers 30 percent below normal, leading to passenger crowding and delays. This reduced capacity may persist until the signals are fully fixed, but it’s no reason to also cut midday service.


Before the derailment of June 11, the Red Line averaged 13.9 trains per hour at rush hour, and 9 trains per hour at off-peak times. On average, there is a train at rush hour every 4 minutes and 20 seconds and every 6 minutes and 40 seconds outside of the peak (the times are twice as long for the Ashmont and Braintree branches), which compares well to the public schedule. The actual operation varies somewhat based on operating conditions and train spacing.

Since the derailment, train service on the Red Line has been affected by slower speeds, signal restrictions, and labor issues. The first two are hard to solve. Until the signals are fixed, the T will have to run slower trains with manual signals, which limits line throughput. The labor issue, which affects midday service, requires no technical of infrastructure repair, but the T and the Carmen’s Union would have to join together for the good of the general public.

The following is our examination of the data for each of these categories.

Because of the speed restrictions in the area where the now-destroyed signal systems controlled train movements, the running times of the trains have increased. Since the T only has a finite number of trains and operators, having slower trains means that they can’t run as frequently.

Let’s imagine a simplified line where the end-to-end running time is 60 minutes (including the time it takes to turn the train around at the end of the line) and service is provided every five minutes. The math is that this line would require 24 trains to operate: the “cycle time” of the line—the time it takes for a train to run from one end of the line to the other, and then turn back—is 120 minutes, and 120/5 = 24. Now imagine there is a delay of 12 minutes added to the service because of a speed restriction in each direction. The cycle time now increases to 144 minutes, yet there are still only 24 trains available for service. This would mean that with the same number of trains, less service would be provided: in this case, the wait between trains would increase from five minutes to six.

This is basically what has happened to the Red Line, although because of the branching, the data is a bit more complex. Since the derailment, the average run time for a Red Line train has increased by 20 percent (by about 8 to 11 minutes, depending on which branch), and the increase is relatively constant regardless of which branch the train operates on or whether it is operating at peak or off-peak times.

If this were the limiting factor to providing service, we would expect that the T would be able to run 20 percent fewer trains at rush hour — 11.5 trains per hour. But that’s not the case, since the T is running only 9 to 10 trains per hour at peak hours, a 50 percent drop in service from before the derailment.

The second potential reason for the reduced service is the throughput through areas without signals. The limit to the number of trains which can be run is based on the safety of manual block signaling, where MBTA staff confirm that the passage between two stations is clear before allowing a train to proceed. While harder to illustrate graphically, given the more severe reduction in service, it appears that this throughput issue is why the T is running the anemic post-derailment rush hour service levels on the Red Line.

This does not, however, explain why the T has cut midday and weekend service. Before the crash, the T ran approximately 9.3 trains per hour at off-peak times (a train every 6 minutes 30 seconds, or 13 minutes on the branches). This is about the number of trains the T is running today at rush hour, so it is clear that it is physically possible to move that many trains across the line. However, the T has dramatically cut off-peak service since the crash. Since mid-June, only 6.3 trains per hour have run during off-peak times: a train approximately every 10 minutes from JFK/UMass to Alewife, which means that there is service only every 20 minutes on each of the branches.

Midday, evening, and weekend trains are, therefore, less frequent and more crowded, meaning that passengers at those times are more likely to avoid the Red Line and drive instead, adding to regional congestion and pollution. Midday passengers are also more likely to come from lower-income groups, so this reduction has an outsized effect on this population. Signal throughput should not be the cause of reduced service during the midday. So, why isn’t the T running the normal number of trains during off-peak hours to mirror the service levels they provide at rush hour?

The third potential issue that could affect the number of trains is the availability of labor. The MBTA schedules the work hours of its employees four times per year, each of which is referred to as a “rating.” The current schedule went into effect on June 23, but employees locked in their shifts weeks before.

The issue is not a question of operating cost: the T is, in fact, running significantly fewer service hours today, even taking into account the longer running times because of the crash. Before the derailment, the T ran about 320 train-hours on the Red Line each day. Since then, the number has been more variable, but ranged from about 230 to 290. Still, it is a 10 percent to 30 percent decrease in the hours the trains are run, and the number of operators needed to operate them overall.

The reason that the T is running less midday service on the Red Line stems from the difference between peak and off-peak service levels, but, like Elizabeth Warren would say, there should be a plan for that: reallocate labor to meet the current condition of the line. This has nothing to do with derailments, signal problems, or deferred maintenance, but rather it is because operators choose their work schedules well in advance, and the schedules are based on normal operations. With less overall service, the operators still come to work, and the difference between the amount of labor provided and the amount of labor required is made up by longer breaks and more layover time at terminals. By running fewer trains at rush hour, the T is not actually paying fewer operators but instead using them inefficiently. Providing more and better service, therefore, is not a cost issue.

Here’s the problem. With the current schedule, employees are locked into this rush hour shift when not all are needed. The T needs the flexibility to move operators from peak to midday service, something that the Carmen’s Union should help to provide by asking its members to change their job assignments. In general, schedules are not changed, and small operational changes (due to bus shuttles or other events) are covered by overtime. However, given a months-long change in run times and throughput, the T should have devised a new schedule which reallocated labor from the peak hour to the midday, to provide the right number of employees at all times.

Imagine the following simplified crew schedule.

During the morning, there may be 100 operators who are scheduled to run trains, but there’s only space for 80 trains on the tracks given the signals, so only 80 are required (operators would not likely be used to man signals, which would be run by more experienced staff). During the midday, evening, and weekend, fewer employees are needed. Under normal operations, T might only need 50 operators during those off-peak periods. But because the trains are running slower, the T actually needs more like 65 or 70 to provide normal headways. There are plenty of physical train cars, but because the trains take extra time to get across the line, there aren’t enough operators to drive the extra trains which would be needed to run normal service. To compensate, the T runs less service.

This should be an opportunity for the T and the Carmen’s Union to work together to recalibrate schedules to move some operators from peak schedules, when there are currently more operators than trains, to the midday, when there is more need. By running the same level of service all day, simpler schedules may result. These schedules—with trains running every 6 to 7 minutes all day long—may all but eliminate part-time work and “split shifts,” where, because of peak demand for operators, some work for a few hours in the morning, then spend several hours on duty but not working, and then work a few hours in the evening. If necessary, supervisors and operators could be offered overtime or moved from other lines to maintain service levels.


It’s possible that there is another explanation for this reduced midday service, but unless the T can provide a reasonable, logical explanation as to why service levels are reduced outside of rush hour, we call on the T and the Carmen’s union to work together to increase midday, evening, and weekend service levels to nine trains per hour to match the throughput of rush hour service as soon as possible.

In addition, because of the persistence of the reduced service at rush hour, we ask that the T work with Keolis to reinstate the additional commuter rail trains serving the Red Line corridor and providing extra capacity at rush hour along the Braintree branch of the Red Line.

It has been more than a month since the derailment, and service has not returned to anywhere near normal. The T should be doing everything in its power to provide as close to normal service as possible. With Mayor Walsh writing to the T to ask for more midday service, bringing the Red Line back to normal would be a good place to start.

The following chart shows the number of trains per hour before the derailment (solid red line) and after (dashed red line). The gray dotted line shows the average number of trains the Red Line runs during the morning peak hour, which is more than the normal midday service, yet the actual service provided is far lower. The yellow line shows the approximate staffing required for current operations (it is equal to the number of 130 percent of the number of trains per hour, to account for additional travel time and additional unpredictability). Think of the solid red line as showing the normal supply of labor, and the yellow line as the demand for labor. During the AM and PM peaks, there is more labor supplied than demanded. During the midday, there is more labor demanded than supplied. (Note that data for the evening may be impacted by track work on the Braintree branch.) During the middle of the day, the T is running as many trains as staffing will allow, despite the capacity on the tracks for more. The arrows show times when labor could be changed: reductions in staffing at peak hours could be used to fill the gaps during other times of day.

Meet the Author

Ari Ofsevit

Boston program senior manager/Board member, ITDP/TransitMatters
Meet the Author

Chris Friend

Member/data scientist, TransitMatters
Collective bargaining and organized labor exist to protect employees; they are not meant to establish barriers to providing critical transit and rail services to people who need reliable and efficient access to jobs, healthcare, and other destinations. This access is not limited to peak rush hour times, and the T and the Carmen’s Union must act in a manner that reflects the larger public interest. Finally, the T should have a standard protocol to work with labor in the future to assure that schedules are not reduced beyond what is absolutely necessary for safety, and a streamlined procedure to run as close to normal service as possible.

Ari Ofsevit is a member of the TransitMatters Board; Chris Friend is a data scientist and TransitMatters member.