clock menu more-arrow no yes

The real reason New York City can’t make the trains run on time

Don’t buy the excuses about overcrowding.

New York’s subway is in crisis. After years of growing ridership, use of the system took a dip in 2016, with further declines this year. Last month saw a train sit in the tunnel in sweltering heat for 45 minutes and a derailment with dozens of injuries. The media describes the situation as hell (Slate), a meltdown (Curbed), or a crisis (NBC). Ryan Cooper in the Week and many irate subway riders on Twitter squarely blame Gov. Andrew Cuomo, who controls the state agency responsible for the subway, the Metropolitan Transit Authority (MTA).

The MTA itself says the problems are about overcrowding. But this is an excuse. First, Chicago has had as large a growth in L ridership as New York has on the subway, without the recent breakdowns. An inside source at the MTA explains that by policy, each train delay must be logged with an official cause and that when no cause is formally identified, they often write down “crowding” as a default catchall rather than a real explanation.

The superficial cause of delays is that maintenance and rule changes have made the subway slower, but the MTA has not adjusted the schedules accordingly. The long answer requires an overview of the history of the subway, and why maintenance has always been so fundamental.

Deferred maintenance has been an issue since the subway’s birth

The first rapid transit in New York, the steam-powered Ninth Avenue Elevated, opened in 1868. Within the next two decades, many additional lines opened in New York and in Brooklyn, at the time a separate city. These lines were all built by private companies, intent on making a profit. Some did not — railroad bankruptcies were routine in that era — but most did.

Fares on the elevated railways were fixed by law at 5 cents, about a dollar in today’s money. But in the late 19th century, the United States was perhaps one-tenth as rich as it is today. Consequently, the fare was unaffordable to most of the working class, which stayed on the Lower East Side, where population density kept rising, leading to overcrowding. The neighborhoods that developed around the els, such as Yorkville, were for the most part middle-class. This situation was not acceptable to the reformers within the city elite, who viewed the situation on the Lower East Side as a disaster of public health and public morality; they hoped the immigrant working class of the neighborhood would move to the suburbs, live in single-family houses, and become proper Americans.

The subway opened in 1904, offering much higher speed: 15 mph on the local trains, 24 mph on the express trains. The els were electrified around the same time. The subway's ridership exploded, and in response to crowding, more lines were built in the 1910s and ’20s. The fare was still 5 cents, but rising incomes made the trains more attractive to the working class. The neighborhoods developed around the subway still featured apartment buildings rather than single-family houses, but the buildings were of higher quality than the old tenements — they had fire escapes, indoor plumbing, and better access to natural light. The Lower East Side's population peaked in 1910, and decreased by a factor of about three by 1930. The most pressing social problems that motivated the urban reformers of the 1890s were largely gone.

But unlike the old els, the subway was not profitable. Also unlike the els, the subway was what nowadays we would call a public-private partnership. New York funded subway construction, giving the private companies that ran the Manhattan and Brooklyn els franchises to operate the lines. After World War I, inflation halved the value of the dollar, but the fare remained fixed at 5 cents. A hike to 10 cents was politically unthinkable, given how poor the city's working class still was. In the 1920s, the operating costs of rapid transit in New York were about 10 cents per rider. During the Great Depression, one of the two private operators went into bankruptcy.

The city’s focus at the time was on building more subways in order to compete with the private companies and drive them out of business rather than maintaining the existing system. In the 1930s, it opened three subway trunk lines largely paralleling the existing els, which were removed shortly thereafter. In 1940, it bought the private systems, eventually creating the modern New York City Transit (NYCT) Authority. In 1948 the fare rose to 10 cents, but by then it was not enough. With money always short, maintenance was never a priority. Deferred maintenance, in short, is a problem that dates from the system’s birth rather than from any new changes.

After World War II, expansion ground to a halt. Annual ridership fell from its 1946 peak of 2 billion to 1.3 billion by the 1960s, as the white middle class left for the suburbs and for neighborhoods beyond the subway's reach. Working-class usage suffered from a cycle of fare increases and ridership decline. By the 1950s, the subway raided funding for the construction of the long-planned Second Avenue Subway in order to keep the system running. But it wasn’t enough.

In the 1970s, decades of deferred maintenance finally caught up to the system. The city's fiscal crisis meant it could not fund any capital projects. The NYCT had, by this point, been subsumed into the MTA, a state authority, but the rest of the state was not doing much better (even the suburbs saw population decline in the 1970s). Ridership dipped below 1 billion. The subway's mean distance between failures (MDBF) was down to 6,000 miles in 1980, and long segments, suffering from deteriorating fixed plant, were under orders to operate at just 10 mph. The Sea Beach Line, carrying the N Train in Brooklyn, had not received any overhaul since it opened in 1915.

The great slowdown

Irritating as today’s problems are, they represent an improvement from the true bad old days of the 1970s. Starting in the 1980s, the MTA invested in long-term maintenance — the “fix it first” approach that many experts now endorse for highways. The MTA began a multi-decade project to reach state of good repair (SOGR), in which there would be no more maintenance backlog and no need to engage in maintenance beyond normal repairs.

This coincided with the MTA’s belated adoption of long-term financial planning, with five-year capital plans, focusing largely on SOGR and normal replacement, and a little money left over for expansion. It scaled back its rolling stock plans: The trains it bought in the 1970s were ambitious (they were supposed to be capable of higher speed) but turned out to be lemons, whereas those it has bought since the 1980s have been conservative, missing out on developments like open gangways that are routine in most subway systems.

By 2010, the MDBF was up to around 170,000 miles. Most slow restrictions had been lifted. Subway ridership was up to about 1.7 billion as a functioning system paired with a return to citywide population growth. The Second Avenue Subway, chopped into four phases due to high construction costs, was under construction, and the first phase would open at the beginning of 2017. Things were looking up.

But even under improved conditions, the trains ran more slowly than they had at the dawn of the 20th century. The oldest locals were back to about 15 mph on average, but the expresses between Grand Central and Brooklyn Bridge, scheduled to take eight minutes in 1906, was only scheduled to achieve this speed early in the morning and late in the evening, running two minutes slower at more normal times. Some of this involved crowding — more passengers mean it takes longer to get people on and off the trains — but mostly it was new slow restrictions.

In 1991, an intoxicated train driver sped through a switch at Union Square, leading to a derailment at 50 mph. Five passengers died (the most of any subway accident in New York since 1928), and 161 passengers were injured.

Four years later, signal failure caused one train to rear-end another on the Williamsburg Bridge, the fourth such collision in two years. The driver died, and several dozen passengers were injured. The train was going at 36 mph through a signal designed for trains whose maximum speed was 28 mph; even with improved braking, the stopping distance was too long. The MTA said the driver could have seen the train ahead, but the driver had been too fatigued. The NYCT responded by slapping a 25 mph speed restriction on all bridges, building on the findings after the Union Square derailment.

After the accident, the NYCT also reduced the acceleration rates of the subway trains to match the rates of the older trains. Subways and regional trains around the world accelerate at 2.5 mph per second or even faster, but in New York acceleration is restricted to about 1 mph per second.

Unfortunately, these modifications to the operating procedure made the old signals less reliable. An MTA source explains that the NYCT did not always recalibrate the signals correctly for the reduced speed. Consequently, train drivers don't trust the signals and go even slower for safety. Train drivers have internalized this practice, and even when the signals are reliable they often go slower than they could, to give themselves a safety margin to avoid crossing a stop signal. Deceleration is supposed to still be quick, 3 mph per second, but in practice it seems to be closer to 2 mph per second. Reporting a problem with the signals is so difficult that the NYCT developed a culture of ignoring it until there's a crisis.

The MTA capital plan includes money to address this with partial automation known as “communication-based train control” (CBTC), to improve reliability and capacity and permit trains to go at the maximum signaled speed, at the maximum acceleration they are capable of. The only line with CBTC, the L Train, is among the most overcrowded but has the fewest delays. But the construction costs of CBTC are extreme. The 7 Train, about 10.5 miles long, has just installed CBTC, for $550 million in trackside infrastructure alone. At the same time, Paris fully automated Metro Line 1, of about the same length as New York's 7 train, for €100 million, or about $125 million, and is doing the same for Metro Line 4, about 7.5 miles long, for €150 million.

New York also has rules for speed restrictions next to work zones. Unlike nearly every other subway system in the world, New York runs trains 24/7. Most systems perform maintenance at night, when the trains aren’t running. Since New York’s trains are always running, they do maintenance during the day, shutting down one track at a time and imposing special slow orders on trains running adjacent to work zones. These rules were tightened in 2003, requiring more slowdowns, with additional tightening in 2008. The new rules took years to be fully implemented, so some slow orders only began recently.

Old trains mean further slowdowns

The NYCT entered this decade in what appeared to be a promising condition. The trains were slower than ideal, but they were reliable. Ridership was rising, and the maintenance backlog was shrinking. Casual observers expected steadily improving service levels to match the city’s growing population and rising economic fortunes. And the good news is that there is no longer much deferred maintenance of infrastructure. The MTA is investing more in track repairs and in CBTC on more lines. This does, however, require more work zones slowing down the trains.

But there is some deferred maintenance of rolling stock. MDBF is trending down, and in 2016 it was 113,000 miles, down by a third from 2011, when it was 172,000. The oldest train class in the system, the 52-year-old R32, has the lowest MDBF in the system, about 32,000 miles, down from 136,000 in 2005 and 57,000 in 2011. This is not just an artifact of its age: The NYCT deferred maintenance expecting new trains to replace the R32s, but the rolling stock order was delayed and will only enter service this year. Even then, the NYCT will not be able to retire the R32s. One line, the L, will be shut down for long-term repairs in 2019-’20, and the NYCT will not be able to redeploy its rolling stock to other lines, which means it will need every available train set elsewhere to absorb the L's ridership.

This means the trains are running more slowly. By itself, that shouldn’t mean less reliability. The MDBF, while declining, is still high enough that breakdowns should not be a regular occurrence. But on-time performance has been slipping largely because the MTA’s scheduling practices are flawed and dishonest.

New York doesn’t tell the truth about the schedule

The slowed-down subway can’t meet its schedule largely because the schedules themselves aren’t done correctly or honestly. The reasons for this fall into four buckets:

1) Labor. Crew get paid for every minute the train is scheduled to run in regular service. If a train is canceled due to track work, the NYCT still has to pay the crew for the hours they would have driven it. If it arrives at the terminal early, the NYCT has to pay the crew as if it were on time. If it arrives late, the crew are eligible for overtime only if it is the end of their shift. For the most part, the incentive is not to adjust the schedule for recent slowdowns.

2) Losing face. The MTA planning department is cautious and constantly fears board intervention, political intervention, or poor PR. The planners have not adjusted the schedules for the recent slowdowns, for fear of embarrassment. With a fixed number of drivers and trains, slightly lower speed means slightly lower frequency. If the planners did adjust the schedules, it would be a visible service cut, and the MTA board would ask why.

3) Poor metrics for schedule reliability. The MTA has two metrics for punctuality: on-time performance (OTP) at the end of each train run and wait assessment (WA), measuring the gap between two successive trains. Since subway riders do not look at schedules, even headways between trains matter more than the schedule, leading to a growing emphasis on WA. This measures adherence to the schedule, relative to the headway, even when the schedule does not have even headways, but the dispatchers still sometimes hold trains to smooth out delays.

However, per an internal NYCT document, about half of the overall delays as experienced by passengers occur on a moving train, and half occur due to excess waiting time, each about 90 seconds. WA is also averaged across all lines, without weighting by ridership, so delays on the busiest lines count as much as on the least busy ones. Zak Accuardi, a senior analyst at the national think tank TransitCenter, proposed doing away with WA entirely and replacing it with a metric combining excess wait time and excess train travel time.

4) The highly branched nature of the subway. In Paris, there are 16 Metro lines, none of which share track with any other. In New York, four subway lines have dedicated tracks (1, 6, 7, L), four more share only with one another (2, 3, 4, 5), and the remaining 13 all share tracks at various points. This means delays on one line propagate to other lines. The timetables already provide for awkward service gaps on some lines, but if trains do not adhere to a precise schedule, most of the system can experience delays. Despite this, if a train is delayed due to an incident, dispatchers will attempt to delay the trains ahead of it, in order to smooth out the WA metric, ultimately leading to both more service gaps and longer train trips. The Daily News covered this in May, showing how the practice of holding trains was making reliability even worse.

The MTA needs to get real

Despite the growing chorus of complaints, in practice the MTA and NYCT are running on autopilot. The NYCT is not making big changes; the MTA is making little effort to get its constituent agencies to cooperate, including the NYCT but also the two commuter rail operators.

With timidity ruling the day, the MTA never really thought about long-term service. The five-year capital plans do include some long-term projects for infrastructure investment, but there is less long-term thinking about schedules and service plans. The NYCT still thinks in terms of going from crisis to crisis. It's this short-term thinking that led it to defer maintenance on the R32 trains. In public relations, the same short-term thinking led the planners to keep pretending the older schedules were fine. Rather than admit to problems and change the schedule, they kept stressing the system until it couldn't take it anymore.

Nor is the political system really interesting in managing the MTA better. Gov. Cuomo is treating the subway more as a source of political capital than as a place in which to invest his political capital. His proposals ignore the banal day-to-day management of the MTA. NYCT managers are willing to do what he asks, but he is asking for pie-in-the-sky tech industry fixes rather than more honest scheduling.

The crisis of the 1970s taught the MTA to be conservative about system expansion, equipment, and speed. This worked in the 1980s and ’90s, but more recently it has created its own problems, including delays, unreliable schedules, and slow trains. Management is overly cautious, which created a new set of problems, distinct from the ones that plagued the system in the era of deferred maintenance.

There is only so much short-term thinking the subway can take. In this crisis, what the MTA needs is the willingness to move forward and get faster, and the internal expertise to be able to implement it. That means starting with an honest assessment about what is a realistic schedule for the present system to meet, so that trains can be reliable if they can’t be fast. Next, authorities need to take a rigorous, detailed look at how to improve train speed — ideally featuring experts involved in the management of large foreign subway systems who can provide an informed outside assessment of whether the MTA’s current safety practices are genuinely justified.

Signal upgrades are clearly a good idea, but they could be accomplished much more quickly if the MTA could manage to get this work done for “only” twice what Paris spends rather than four or five times as much. Union work rules should not create an incentive for misleading schedules, and official reliability metrics should be designed to capture riders’ actual experience, not to put a good face on things.

Most of all, New York needs an end to the era of generic excuses about crowding, and a set of agency and political leaders who are instead willing to level with the public about the state of the subway and begin offering recommendations for what it really needs — painstaking operational improvements, not high-tech quick fixes.

Sign up for the newsletter Sign up for The Weeds

Get our essential policy newsletter delivered Fridays.