‘The Trains Are Slower Because They Slowed the Trains Down’

Internal MTA documents show everything we thought we knew about subway delays was wrong


In the summer of 2014, New York City Transit intern Philip Betheil was finishing up his master’s in urban planning at Columbia University when his boss, David Greenberger, gave him a project. The two worked for NYCT’s operations planning division, and Greenberger tasked Betheil with looking into an arcane bit of subway minutiae called signal modifications and what effect they had on train service. They worked on the report on and off over that summer, tossing more than a dozen drafts back and forth.

In August of 2014, after Betheil’s internship had ended, the draft report languished in the organization’s digital innards. (Betheil declined to comment on the record for this article. Greenberger did not respond to a request for comment.)

But now, more than three years later, the report, which was obtained by the Village Voice along with other internal documents, provides a radically different explanation for the subway’s declining performance than the one that MTA leadership has given the public. The root cause of the subway system’s decay, it turns out, isn’t budget cuts or overcrowding — rather, the collapse of the subway system appears to have been primarily self-inflicted by the authority itself, in response to a single accident two decades ago that set the transit system on a path to disaster.

Moreover, these internal documents suggest that much of what the MTA is doing to fix the subways, including the authority’s $836 million Subway Action Plan, is not addressing the bulk of the delays that are plaguing the city’s transit system. And only now has the subway’s leadership, since the recent hiring of New York City Transit president Andy Byford, begun to seriously consider its own role.

“It’s not that complicated,” a source with direct knowledge of the situation who asked not to be identified told the Village Voice. “The trains are slower because they slowed the trains down.”


In recent years, the MTA’s preferred narrative for New York City’s subway problems has gone like this: Budget cuts by prior governors and mayors over the past two decades have led to deferred maintenance, which in turn caused the condition of key components such as signals, tracks, and cars to erode over time. Eventually, the cumulative impact of all these failing parts reached a critical mass, at which point subway delays began to mushroom out of control. The Subway Action Plan, announced last July by MTA chairman Joe Lhota, was designed to address these issues.

Nobody debates that deferred maintenance has had a toll on the subway’s resilience and that the system’s antiquated signals need modernization. But new evidence examined by the Village Voice shows that — though the MTA disputes this — simply replacing old technologies with new ones may not be enough to solve all the subway’s fundamental problems.

The subway’s performance has been steadily deteriorating for many years. The authority’s own internal data shows that delays due to “incidents,” such as broken signals and tracks or water damage, have only marginally increased since 2012. But there is one type of delay that’s gotten exponentially worse during that time: a catchall category blandly titled “insufficient capacity, excess dwell, unknown,” which captures every delay without an obvious cause. From January 2012 to December 2017, these delays increased by a whopping 1,190 percent — from 105 per weekday to 1,355. In December, one out of every six trains run across the entire system experienced such a delay. The increase has been steady and uninterrupted over the past six years:

In other words, as the rest of the system has remained relatively stable, this catchall category has worsened to the point that it’s gone from a minor problem to the single biggest cause of delays in the system.

This, almost certainly, is where the other primary MTA narrative developed, that of “overcrowding”: Historically high ridership has resulted in crowds that are keeping trains in stations longer. During the press conference announcing the Subway Action Plan last July, Lhota called the “increasing volume of passengers” the “main issue” facing the subways. The previous month, the New York Times had run a major feature titled “Every New York City Subway Line is Getting Worse. Here’s Why,” which bluntly stated, “The major cause of subway delays is a factor that basically did not exist fifteen years ago: overcrowding. The subway is a victim of its own success and the city’s resurgence. Large crowds slow down trains, which creates more crowding in a vicious circle that takes hours to unwind during every rush.” The main source for this assertion was the data included in the MTA’s monthly board books.

Many — including this reporter — bought the MTA’s narrative (although some didn’t) because the agency’s own publicly released data reflected it (and because the subways were, in fact, frequently crowded). Meanwhile, the MTA’s performance reports in its board books, a key source of information for reporters and the authority’s board members, had changed the title of the internally used “insufficient capacity, excess dwell, unknown” category to simply “over crowding,” making the same numbers seem far more conclusive. (An MTA spokesperson, who refused to be quoted for this article, told the Voice that the reclassification was intended to reduce jargon.)

Even as the subway’s performance tanked, the MTA not only continued to blame overcrowding but touted to its own board members the effectiveness of its proposed remedies. In April 2017, MTA board member Charles Moerdler noted the ever-rising “over crowding” figures and asked what the MTA was doing about it. Wynton Habersham, senior vice president of the NYCT’s Department of Subways, answered that his office was having “some success” with platform controllers who had reduced the amount of time trains remained in the station “a good amount.” In fact, platform controllers — known to most straphangers as the MTA employees waving flashlights — had first been deployed two years prior to Moerdler’s inquiry. In those two years, even though overall ridership fell, delays attributed to overcrowding still increased by more than 50 percent.

The MTA continues to defend the use of platform controllers’ role in reducing delays. An MTA spokesperson told the Voice that platform controllers have helped improve or stabilize dwell times, as well as prevent delays in other ways, such as by tending to medical emergencies or troubleshooting problems. (The authority does not publicly release data on station dwell times.)

A subsequent Times investigation pointed out how delays attributed to overcrowding continued to rise even as fewer people took the subway, adding that the MTA had a historical pattern of muddling delay records in an effort to “avoid blame for the system’s problems.” Overcrowding’s predecessor was “supplement schedule” — a category that captured nearly 80 percent of all delays from September 2009 to May 2010, according to the Times, because “whenever maintenance work in the system caused a scheduling change, virtually all delays were put under this label, regardless of their cause.”

Nevertheless, something is causing almost 1,400 trains to run late every single weekday. As it happens, the never-released 2014 report explains one of the major causes. And it has nothing to do with overcrowding.


In 1995, a Manhattan-bound J train crossing the Williamsburg Bridge rear-ended an M train that was stopped on the bridge, killing the J train operator and injuring more than fifty passengers. The National Transportation and Safety Board investigation placed most of the blame on the J train operator, who the NTSB suspected had been asleep. But the NTSB also identified potential issues with the signal system that contributed to the accident, which it found didn’t guarantee train operators enough time to apply the emergency brakes even when awake.

“They slowed the trains down after the Williamsburg Bridge crash,” a veteran train operator who asked not to be identified told the Village Voice. “The MTA said the train was going too fast for the signal system.” As a result, the MTA, quite literally, slowed all the trains down, issuing a bulletin informing employees in April 1996 that their propulsion systems would be modified so they could achieve a maximum speed of 40 miles per hour, down from the previous high of 50 to 55 miles per hour on a flat grade.

But the MTA didn’t stop there, internal documents show. One of the NTSB’s safety recommendations was to set speed limits. As a result, the MTA began a still-ongoing process of changing the way many signals work to meet modern safety standards. Previously, trains would encounter green lights if the track ahead was clear; if not, a red light would trip the train’s brakes if it tried to cross that section of track, stopping it a safe distance from the train ahead. (Restarting the train after it is tripped can take anywhere from one to ten minutes, and the incident gets noted in the operator’s performance file, according to MTA sources.) Under the new program, generally referred to as signal modifications, the brakes would be tripped based not only on whether the track ahead was vacant, but also on the train’s speed.

An MTA spokesperson said safety is the authority’s primary goal and that measures were put in place to ensure the Williamsburg Bridge crash never happened again.

NYCT had predicted that the signal modifications would only marginally affect run times. But the 2014 study — the first time the authority had attempted to analyze the impact of any of the revamped signals, using its improved data system — found 2,851 lost total passenger hours per weekday could be attributed to thirteen modified signals alone. That was almost double the predicted impact; for comparison, the modifications of those thirteen signals alone created 5 percent as much lost time as that experienced by riders of the entire London Underground on its average day. (The MTA declined to comment directly on the 2014 study because it was never finalized.)

NYCT’s estimates were so off in part because they didn’t account for a human element. The most problematic of the newly modified signals were “one-shot timers,” so called because the operator has only one chance to meet the posted speed limit. One-shot timers are easier and cheaper to install, say MTA sources; the more “shots” the operators have to get under the required speed, the more timing mechanisms have to be installed across a longer portion of track. (An MTA spokesperson disputed this, characterizing the decision to install one- or two-shot timers as safety-related.) But the consequences of going over the speed limit are high — the train is stopped, and the operator gets penalized — so many operators now opt to go well below the posted speed limit just to be on the safe side.

“The grade time signals force us to operate slower, and because they have been installing them gradually, the subways have been slowing down gradually,” the train operator told the Voice. “I used to be able to go from 125th Street to 59th Street on the A line in seven minutes. Now it takes around nine minutes!”

In a private Facebook group for train operators, others note the same frustrations about the signal modifications — especially the one-shot timers. “All these one shot timers they put in over the past few yrs are ridiculous,” one veteran operator commented on the Facebook group last year. Several operators have also expressed suspicion that some of the signals and train speedometers are miscalibrated, further exacerbating the problem. Another commented that she was told 10 mile-per-hour speed limits “means 5-7, especially with b.o. [“bad order,” or broken] speedometers.”

An MTA spokesperson insisted that these allegations by train operators were untrue.

Though the report only studied these thirteen specific signals, it warned that “it is likely that cumulative running time effects will only further increase” as NYCT rolled out the signal modifications to the entire system. (A 2012 MTA document stated that more than 1,200 signals had been modified, with plans to further modify approximately 500 more, though not all the modifications affected run times the same way.) And the report noted that although the delays “represent fairly minor impacts,” they could “potentially affect timetables and train requirements if sustained throughout an entire line, generating unanticipated operating and capital costs.”

NYCT has also imposed increasingly stringent speed restrictions around track work. A Daily News report from last May detailed how trains are required to go a mere 10 miles per hour for a third of a mile before reaching the actual work site. The agency has also expanded speed restrictions to adjacent tracks, even if no work is being done on them. As a result, delays resulting from track crews and planned maintenance on weekdays have increased 69 percent from 2012, according to Daily News reports and confirmed by the Village Voice.

“If we were delayed by something like a sick passenger or a signal problem, obviously I would know,” the train operator who spoke to the Voice said. “But usually I’m late because of many smaller problems.”

Sometimes, he says, his train will get to a junction point with another line on schedule but will have to wait for a train that’s running late to pass in front of him; other times, a conductor slow to close the train doors will result in the train dwelling longer than it should, or he’ll be stuck at a station trying to speak to a dispatcher on the MTA’s balky radio system. On certain lines, he says, “I’m always late because the grade time signals slow me down, but the schedule doesn’t provide enough running time for me to make the trip on time.”

The MTA says it regularly adjusts timetables, most recently in December for the 2, 3, 4, and 5 lines. According to the most recent MTA board book, the 2, 4, 5, 6, A, B, D, and F were on time for less than 50 percent of their weekday trips in December. Systemwide weekday on-time performance was 62.9 percent.

“All of the little things like this add up to a lot of lost time,” the operator said. “When dispatchers see that we’re late because of these little things, they usually just put down ‘overcrowding.’ ” (The MTA rejected the suggestion that this is standard practice, adding all train dispatchers and tower operators are required to write down the accurate reason for a delay.)

The same Daily News article reported that in 2015, then-NYCT president (now managing director of the MTA) Ronnie Hakim inquired why the word “speed” was included in a Department of Subways poster with the slogan “Safety, Service, Speed, Smiles.” When the Daily News followed up with her about the remark, Hakim replied, “It would concern me that people would feel pressure to be faster rather than to be safer.”

But the train operator who spoke to the Voice and has been running trains since the 1980s sees this as a false dichotomy, one that has cost riders dearly. “It feels like the MTA doesn’t care about speed anymore,” he lamented. “They always act like speed is unsafe, but you can run trains quickly without sacrificing safety if you have competent management.”

An MTA spokesperson called this statement a mischaracterization of its mission, and said it was unfair to the hard work that employees do to make the subway system work.

Benjamin Kabak, who writes the subway-focused blog Second Ave. Sagas and has been one of the few transit advocates raising concerns about the subway’s slowdown, agrees with the operator’s assessment. “I don’t think they’re wrong in citing safety,” Kabak says of MTA management, “but I think there’s a question as to whether their reaction is commensurate to the problem.” He doesn’t see a safety benefit to the vast majority of timers installed around the system. As an example, Kabak recently noticed the 6 train crawling between 51st Street and Grand Central on a regular basis. He suspects a timer has recently been installed there.

“There’s really no explanation as to why they felt the need to slow down the train there,” Kabak says. “It seems to me like they’ve really just gone overboard with them.”

For an example of what can be achieved with a focus on both speed and safety, look to Asia. For just one example, the Tokyo metro system has 8 million daily riders — 2 million more than New York — and viral video–worthy crowding issues, yet maintains an on-time performance around 99 percent and a stellar safety record. A key reason the system is globally known for its obsessive punctuality is its early adoption of automated train control technologies that make the system both safer and more efficient. While the MTA was rolling out timers to slow trains down, Tokyo was installing signals that use automation to let trains go as fast as possible, the very kind of signals New York is struggling to incorporate today. The most punctual subways in the world care not only that trains are staying under an unsafe speed, but also that they’re going as close to that speed as possible at all times.

Of course, similar types of semi-automated train control have helped the New York subway as well. The L is the only line that currently has communications-based train control, or CBTC, which tells train operators precisely how fast trains ought to be going at all times, taking into account both safety requirements and efficient operations. Mostly thanks to this technology, the L’s weekday on-time performance generally hovers around 90 percent, or about 30 percent higher than the system average, despite being one of the most heavily used lines in the system.

But as anyone who has ridden the L knows, CBTC is not a cure-all. It can help eliminate the human element that keeps subway trains operating below the posted speed limits and reduce headways, but there is still the issue of the unnecessary speed limits themselves. Moreover, even the most optimistic projections forecast it will take decades to install modern signals for the rest of the system; in the meantime, resetting existing timers to allow for higher speeds can help reduce delays. And even a modern signaling system with the same overly cautious speed restrictions would still result in a suboptimal subway system.

Kabak sums up the problem: “If you’re running trains at speeds that are lower than what the system capacity was designed for, then you’re losing capacity, and that’s a choice that you’ve made. If that’s a choice that you’ve made, then you have to prove to people why you’ve made that choice. And if it’s a question of a hypothetical safety situation that doesn’t come into play, then you kind of have to question how that analysis has been reached.”


Just like “supplement schedule,” the overcrowding theory has outlived its useful life. Starting with the October 2017 board materials, the category was finally modified to “over crowding/insufficient capacity/other,” more closely reflecting the nebulousness of the internal definition. Similarly, recently hired New York City Transit president Andy Byford, who has held senior positions in the London, Sydney, and Toronto mass transit agencies, asserted at the last NYCT meeting, “If we’re to truly improve the service that we offer, you have to get to the underlying root cause.… So therefore I don’t want to just see ‘overcrowding.’ I want to see what caused that overcrowding, what was the absolute underlying root cause.”

In an email to the Village Voice, Byford acknowledged that “changes made to the signal system (in response to a crash in 1995) have undoubtedly had an impact on subway capacity,” and added that NYCT leadership has already met to begin reviewing the issue. “We are studying the impact and what was done to see if adjustments can be made while still maintaining the safety benefit these changes (and more onerous flagging arrangements) were brought in to address. This renewed scrutiny is part of my drive to properly understand — then tackle — root cause. As I have already said, the real fix is renewed signal systems and that will be the anchor behind my Corporate Plan that I am currently working on.”

Back in the train operator Facebook group, several members remain irked by a recently installed one-shot timer on the northbound 4/5 track between Franklin and Atlantic avenues in Brooklyn. “That timer at Atlantic is the youngest one,” an operator commented. Trains used to be able to go more than 40 miles per hour between Franklin and Atlantic, but now much of the track is restricted to below 25 miles per hour, and a mere 10 miles per hour approaching the station thanks to new safety rules, according to multiple MTA employees, although at least some operators have been instructed to go even slower than that.

Because the timer is new, one operator complained, dispatchers at Rail Control Center may not have known about it and therefore asked if there were any problems. “I would’ve been like yeah I got a problem,” another operator replied. “Y’all put this busted ass timer here.”

The operator who spoke to the Voice said his biggest frustration on a day-to-day basis is the same as every straphanger’s: the Sisyphean task of keeping the train on schedule. “I try to keep my train moving,” he insisted, “but there’s a lot working against me.”

Next: If the MTA has been trying to fix the wrong problems, what can and should it be doing to get the subways up and running again?