Maintenance, Process, and the Future of the Off-Peak Subway

Imagine a world where the R train ran every eight minutes on weekends, where express service on the 7 line did not cease during middays, where you could catch a B train in Manhattan late into the night, and where the Saturday Rockaway A train qualified as frequent transit. Sounds nice, right? This fantastical universe once existed: it was the year 2006. Since then, non-peak ridership on New York’s subway has increased rapidly — and service has been slashed significantly. Understanding how and why this happened is not only important for our historical appreciation for the forces shaping New York’s troubled transit network, but is also critical to efforts facing down our climate crisis: we must make the off-peak subway work again.

The story of the subway’s comeback since the dark days of the 1980s is, unequivocally, one of off-peak ridership growth. On weekdays (as Alon Levy often notes), virtually all the growth in Manhattan core subway use has occurred in hours outside the traditional peak. Weekend ridership data for years that far back is hard to come by, yet is impressive in the time periods available. From 1998 to 2014, weekend ridership grew by an astonishing 70 percent, compared to 41 percent growth during weekdays. These trends reflect a fundamental shift in use patterns towards discretionary trips, and towards non-traditional working hours: through these decades, the subway became a more flexibly used and deeply embedded element of peoples’ daily lives.

Continuing the trend away from a peak commute oriented subway will be central to successful efforts to decarbonize New York, promote transit equity, and rebuild transit ridership after the end of the pandemic. Yet today, the subway’s off-peak renaissance faces severe headwinds. Despite this remarkable growth, and despite the never-ending drumbeat for increased service, only two of the seventeen subway lines for which complete data are available actually saw their weekend frequencies increase from 2000 to 2021. On the rest of the system, frequencies have remained constant or decreased; on many lines, the scheduled time between trains has grown by 50 percent. Driving this service reduction — and ones of comparable scale during weekday evenings — is not some malicious plan by bureaucrats or politicians to deprive New Yorkers of essential transit service, or even bad service level policy on the part of the MTA. Rather, at the heart of the system’s off peak service struggles lies the way we do maintenance.

Where We’ve Been

The route to our troubled present can be traced all the way back to the system investment decisions of the 1950s, but one of its most important threads picks up in 2007, on a MTA-TWU working group called the Track Safety Task Force. Following the deaths of two workers struck by trains, the agency undertook a full revision of its roadway worker protection rules. The committee issued a number of recommendations, but — as Adam Pearce has previously reported in the Times — key among them was one that trains running adjacent to work zones should slow to 10 miles per hour as they passed, rather than running at track speed as they had previously. This had important ramifications for service quality during the off-peak hours when trackwork takes place, adding two additional levels to the basic disruption that is a service change. Most obviously, it made the trains significantly slower. It also reduced the number of trains you could run on a corridor when doing work. As explained in the graphic below, most subway signals keep trains a fixed distance apart. When trains slow significantly below this maximum design speed, however, the distances by which they are separated (generally) do not change, increasing headway, or the time between trains. Where a corridor with a track out of service may once have been able to handle 20 or 25 trains per hour, the new rules forced service levels down to 15 or 18.

If the agency limited the amount of work taking place on the tracks, the impacts of these new rules could have been controlled; perhaps work would impact service on some nights and a handful of weekends each year. But the trend over the past decade in the subway has been — strongly — in the opposite direction. Aging infrastructure and a long backlog of repair work has meant steadily increasing planned disruption through the late aughts and 2010s. Though the agency recognized the problems associated with increasing work loads, and made efforts to consolidate work in 2011-15, their emphasis on finding less impactful ways to maintain was lost in the response to the reliability crisis of 2016-18. A key element of the somewhat misguided Subway Action Plan was a brute force ramp-up of various cleaning and repair programs, and, critically, a decision to make all weeknight service changes begin at 10pm, rather than at midnight. Though the Subway Action Plan’s work has largely concluded, changes from the era have remained sticky: maintenance loads on the system remain high, and service changes still begin at 10; the crisis-driven decision to deemphasize service in favor of renewal has become a part of the system’s institutional fabric.

On a longer timescale than the Subway Action Plan and the increasing number of maintenance outages, capital improvement projects have played a similarly important role in disrupting weekend subway service. Thanks to their greater complexity and intensity, these projects often disrupt night and weekend subway service for months at a time — and are rarely coordinated to reduce impacts. For example, a structural project on the 2, 3, 4 and 5 lines in Brooklyn set to impact trains on the corridor for 33 months began last October. This comes after a year of intermittent service changes related to elevator installation, and only two years after the agency completed its reconstruction of the Clark Street Tunnel and associated suspension of 2/3 train service. Other corridors have had it worse. In Astoria, NYCT undertook a slew of projects from the summer of 2017 to 2020, each of which had significant weekend and midday service impacts. The agency did most of the projects sequentially, creating an essentially unbroken chain of midday and weekend diversions over three years. One would think that Astoria might have some respite after these improvements, but the spring and summer of 2021 has brought only more disruption as track replacement work plods along. Astoria may lie towards the more impacted end of the spectrum, but no corridor has remained unaffected by these service changes. Not only have virtually all lines seen some capital work over the past decade, but the subway system is a fundamentally interconnected entity: delays and frequency reductions stemming from a project in one borough will almost always translate into problems for another.

The net impact of these initiatives has been a significant reduction in off peak subway service levels and ridership. As workloads increased, New York City Transit began cutting service to reflect new realities across the network. The first such cuts came in 2010, when most letter lines moved from an 8 to a 10 minute headway, and the 1 train went from a 6 to an 8. In the last five years, this effort has been followed with further cuts: today, most letter lines run every 12 minutes on weekends, and evening service has been significantly reduced to conform with the earlier service change start times instituted under the Subway Action Plan. These cuts hurt riders doubly: not only did they increase waits for trains, but, in their concurrence with increasing levels of service disruption, they did so in a world where riders often had to transfer many times to complete a trip.

Yet to see these cuts as irresponsible somewhat misses the point. In each of these cases, NYCT’s service decisions were driven by the fact that increased levels of work under new worker protection rules forced off peak service reductions by temporary schedule anyway, so having a base schedule with higher frequencies essentially just made for increased crewing costs with little rider benefit. While an essentially sound decision within our existing maintenance paradigm, this pathway to service cuts should highlight to us the power of maintenance to shape service. The agency has publicly acknowledged that the frequency reductions undertaken for projects reduce service to levels below those called for by their own frequency guidelines; the path of causality here runs not in a frequency-ridership spiral, but instead from maintenance to inutility.

In the early 2010s, it seemed that, despite these headwinds, there might be hope for off-peak subway ridership. From 2010 to 2014, system ridership gains were led by impressive increases during evenings, early mornings, and weekends. This trend predated the advent of 10pm work start times in 2017, but at least on weekends, it seemed riders were willing to tolerate some moderate level of disruption and infrequency — if the trains still, basically, worked.

However, starting in 2014, the system began showing signs of distress. With diversions increasing and the system’s operating environment deteriorating, weekend OTP fell below 80 percent in 2015 for the first time in the decade. Off-peak ridership fell with it. The subway system scored small ridership gains between 2014 and 2016, but the upwards aggregate line masks a story of divergent fortunes. Overnight and weekend ridership began to tumble in those years; the system was being pulled along by its peaks.

After 2016, the bottom truly fell out. We all have read the horror stories of the subway’s reliability crisis, of the slow trains, the incredible incidents, the endless and unknowable waits. These issues impacted all subway riders, but those using trains during the off peak were hurt doubly. Not only did they suffer the basic operational issues afflicting the trains, but also the service consequences of decisionmakers’ efforts to fix them. The most significant ridership losses between 2016 and 2016 were during evening, overnight and weekend hours — those afflicted with this double-crisis of poor reliability and planned disruptions.

Where We Need to Go

Following operational gains made under the leadership of Andy Byford (and before the COVID pandemic), weekend ridership began to bounce back, but we should feel little security in this trend. Setting aside uncertainties about what the post-COVID landscape of transportation demand might look like, the 2020-24 capital program promises to bring only more disruption as the system pursues resignaling, accessibility projects, structural and track upgrades, and the like at a significantly increased pace. Moreover, track access problems are a critical component of the MTA’s cost issues. Poor infrastructure maintenance productivity — likely stemming, in part, from labor requirements associated with the agency’s roadway worker protection rules — significantly increases NYCT’s operating costs, while track access scheduling inefficiencies often increase capital project costs and timelines. Lying at the intersection of the three key functions of an integrated railway operator — service delivery, maintenance, and capital improvement — good track access management is key for the health of any transit system, but for the MTA, improving these processes increasingly seems the system’s greatest challenge. Though it is difficult to offer a prescriptive program given the opacity of track access processes, it is clear that reform efforts demand changes to both when and how the system does its work.

For decades, advocates have called for the MTA to shift away from partial line closures (express trains run local, northbound F trains run via the C, etc) and towards full line-segment shutdowns. Because trains do not interfere with work operations during segment closures, they can often allow projects to move much more quickly than when running partial service. This reduces the number of days suffering disruptions, and can also help moderate project costs. These benefits have made partial closures attractive to many transit operators. Most systems shut down at night for maintenance, but especially in Europe, it is additionally common for operators to close systems for days or weeks to allow concentrated and efficient renewal work. Since 2017, the main trunk of Munich’s S-Bahn has shut for two weekends a year for intensive maintenance work, remaining fully available for train service at all other times. On a larger scale, Paris’s RATP uses much-decreased ridership during the French summer holidays to schedule month-long continuous outages. Madrid’s metro employs a similar system to RATP, closing lines for weeks at a time to undertake heavy maintenance and improvement. Finally, though not part of a regularized system, Chicago is notable for having employed a full-closure approach in its 2013 renovation of its Dan Ryan branch — a strategy which compressed four years’ weekends worth of renewal work into five months of shutdown. Intense, coordinated line closures can be an incredibly effective tool in transit managers’ belt, one New Yorkers’ would be well served to use.

Importing the model to New York will require adaptation to the city’s nuances, however. For one, replicating overnight shutdowns is likely inadvisable in the city, thanks to yard capacity and replacement service constraints. For another, to fully realize the promise of longer closures, the agency will need to thoroughly reform its maintenance support institutions. NYCT’s reliance on small diversions means that it likely does not have the flex capacity in its bus fleet to support replacement service needs for larger closures without expensive hired capacity. And though the agency’s low maintenance productivity means that it probably has sufficient aggregate labor and support resources to handle full closures, these resources would likely need to be redistributed across competencies to maximize work intensity. The agency would not, for example, need nearly as many flaggers for construction projects when revenue trains are not moving around work sites. Adaptations such as these are likely necessary in the long run, but they come in increments: embracing full weekday closures will necessarily be a much more significant effort than an expansion of weekend shutdowns. At least for intermittent maintenance projects, the 80/20 here seems to be a Munich-inspired expansion of the agency’s successful existing weeknight closure program to concentrate work on a few weekends each year. This would avoid the thorny issue of weekday replacement service, allowing monies and institutional energies to be focused on the (potentially more complex) problem of building maintenance and capital support resources sufficient to allow short and intense outages.

Independent of any decision on the time landscape of outages, however, the agency must also focus on improving their outage productivity. Consolidating tasks into shorter windows of unfettered access may yield productivity improvements, but not all of New York’s maintenance problems can be explained away by outage format — especially after considering the success of Japan’s railways. Take Tokyo’s Yamanote Line, the city’s busiest. Despite handling truly incredible volumes of traffic, the corridor went from 1987 to 2019 without a single scheduled outage of train service, accomplishing all repair and renewal work in 3-4 hour overnight windows when the route closed. This is not a feat of consolidated closures, but rather of careful planning to minimize the number of maintenance tasks, and the time each of them takes to perform. Through decades of investment in track components specifically designed to require little maintenance, and the absolute minimization of maintenance-intensive infrastructure elements like switches, Japanese railways have significantly reduced their workloads. What remains to be done is then achieved efficiently within the short overnight windows provided to operators. Such levels of performance are but a mirage on the horizon for New York today, but they demonstrate the need for a commitment to deep process reform in maintenance. NYCT must accompany any changes to the structure of track access with efforts to improve its capacity to piggyback projects, write diversion schedules, coordinate work train movements, minimize the service impacts of adjacent work zones, and otherwise; these details can make or break our efforts to decarbonize off peak transport.


New York is not alone in facing a crisis of maintenance. Washington’s metro ended the 2000s one of the most successful systems in the country — and then fell from grace in the 2010s, as a reliability crisis and a deluge of service diversions and curtailments kneecapped the system’s usability outside the traditional peaks. At the other end of the Northeast Corridor, Boston’s MBTA is facing a large backlog of necessary rapid transit improvement work in a similarly low-productivity environment. And New York’s own commuter railroads are increasingly feeling the maintenance pinch, as their long project timelines and high costs get in the way of reliable, speedy service — let alone any regional rail improvements. As our transit infrastructure ages, and hopefully becomes more intensively and evenly used, the management of track access is set to become an only more significant problem.

But more broadly, the story of New York’s subway maintenance crisis is really just a microcosm of the problems facing American bureaucracies writ large. Spending more money is not alone sufficient to fix ailing institutions; indeed, in this story, investment is actually an antagonist. Nor are these simple stories of institutional malevolence, either. At each turn of this story, NYCT generally made rational decisions to protect workers, control operating costs, and accelerate project timelines. What lacked in was the broader capacity to step back, and rearrange these incredibly complex and objectively difficult tradeoffs between maintenance and service in a way that maximized benefit for riders. In other words, what New York sorely needed is greater institutional capacity, and specifically a sort of institutional capacity that is encouraged and empowered to make major reforms, and apply lessons from the workings of other complex transit systems to New York’s broken present. As countless others have written before me, the American public sector needs a dual revolution: one that gives agencies more money to expand sorely needed services, while also leading them to cultivate better processes. Only thus will we be truly successful in our critically important efforts to cultivate a stronger government for the sake of decarbonization, equity, and social welfare.

5 thoughts on “Maintenance, Process, and the Future of the Off-Peak Subway

  1. “projects often disrupt night and weekend subway service for months at a time — and are rarely coordinated to reduce impacts.”

    This to me is the worst part. Ive seen this on PATH as well. After a full year of closing entire tunnels on weekends, they welcome back full service for something like two weeks before service was disrupted for construction on the draw bridges and signals. After a year of that, they celebrated the work by announcing single tracking due to construction of Harrison Station. It never ends.

    And this is on top of off-peak frequency being cut as well. Nights used to be 30, then 35, now every 40 minutes.And no, the new headways do not account for single tracking disruptions, so a 22 minute trip takes 29 every time.


  2. “At each turn of this story, NYCT genAt each turn of this story, NYCT generally made rational decisions erally made rational decisions…”

    I wouldn’t call them “rational” decisions. “Well-intentioned” is more fitting.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s