By now, New York City commuters are familiar with the wait. We descend from the bitter cold or the stifling heat to find subway platforms teeming with other bodies trying to make it to work on time. Delays ripple through the system, so there’s barely room to squeeze into the next train that arrives.

For years, the Metropolitan Transportation Authority told us that rising ridership and overcrowding were to blame. Yet ridership actually stayed mostly flat from 2013 to 2018 as delays rose, and the authority recently acknowledged that overcrowding was not at fault.

Instead, two decisions made by the M.T.A. years ago — one to slow down trains and another that tried to improve worker safety — appear to have pushed the subway system into its current crisis. And there’s no easy fix.

Just one delay can back up the subway.

To understand how the M.T.A.’s changes have slowed down the trains, let’s look at one delayed train, represented here by a single car.

When a train is delayed, the next one may be forced to slow or stop to maintain a safe amount of space.

Delays can cascade down the line, backing up even more trains.

In the past, the subway could recover more quickly from these cascading delays. But the changes by the M.T.A. have hampered the system’s ability to bounce back.

Worker safety rules and changes to the signal system have made delays last longer.

With the two decisions, the M.T.A. dealt the subway system a blow that may be felt well into the future.

First, the agency decided to increase the amount of space required between trains. It installed or modified hundreds of signals, which regulate train spacing. In that process, signals throughout the system were misconfigured — set up in a way that slowed the trains down even more than officials intended.

Second, the agency adopted new rules for track work that expanded safety zones and increased setup times.

An analysis of internal M.T.A. documents and interviews with system managers and train operators suggest that these two changes removed extra capacity — the ability to run more trains than scheduled — from the subway system. This, on top of years of cost-cutting and deferred spending for maintenance in the 1990s and 2000s, is why the system is no longer able to rebound from disruptions as it once could.

“It’s a conga line of trains all the way down to Brooklyn,” said Kimberly McLaurin, a train operator on the numbered lines who started in 2008. “Any one thing can back up the line.”

Andy Byford, the new president of New York City Transit, “asked for an analysis of the impact of signal modifications on subway schedules” as part of a review announced in January, Jon Weinstein, a spokesman for the M.T.A., the transit agency’s parent organization, said in an email.

Faulty signals force train operators to slow down too much.

After a 1995 collision of two trains on the Williamsburg Bridge killed a train operator and injured more than 50 passengers, the M.T.A. began installing and modifying hundreds of signals to prevent trains from going too fast.

When a train passes over a signal’s switch, a timer starts.

How a signal is supposed to work

Train crosses signal’s switch, triggering timer.

If the train reaches the next segment of track faster than the speed limit allows, the signal will automatically activate the train’s emergency brakes.

Train crosses signal’s switch, triggering timer.

If the train reaches the next segment of track faster than the speed limit allows, the signal will automatically activate the train’s emergency brakes.

Train crosses signal’s switch, triggering timer.

If the train reaches the next segment of track faster than the speed limit allows, the signal will automatically activate the train’s emergency brakes.

The M.T.A. projected that the signal changes would not reduce the number of trains that could pass through a section of track each hour. But this assumed the signals would work properly and that trains would operate at the speed limit.

A good signal allows a train to pass through at the speed limit.

A misconfigured signal will trigger automatic braking even at speeds below the limit.

Some train operators slow down for all signals in case they are passing through a faulty one.

In reality, many signals are poorly maintained and misconfigured, triggering emergency braking at speeds below the listed limit. An unpublished 2014 internal M.T.A. analysis, first reported on by The Village Voice, found that the signal changes caused a significant slowdown, more than the M.T.A. expected.

Train operators face steep penalties after a number of instances of tripping a signal, like losing vacation days or being forced into early retirement.

“If you have two of those type of incidents, I’ve seen people forced into retirement because of something like that,” Ms. McLaurin said, noting that she always approaches the signal timers ready to stop the train, regardless of the posted speed. “I’m not playing with my job,” she added.

“Our instructors are the ones telling us, ‘Don’t go by what’s posted, go 10 to 15 miles lower,’ instead of just having someone fix it or adjust it,” Ms. McLaurin said.

Signals changes caused an immediate slowdown between stations.

4

5

From Wall St. to Bowling Green

4

5

From Fulton St. to Wall St.

25 seconds

40 seconds

20

35

Travel time increased by

12 seconds after signals were

modified in March 2013.

Increased by

3 seconds

in Feb. 2014.

15

30

10

25

5

20

2013

2014

2015

2013

2014

2015

4

5

From Wall St. to Bowling Green

4

5

From Fulton St. to Wall St.

40 seconds

25 seconds

35

20

Travel time increased by

12 seconds after signals

were modified in

March 2013.

30

15

Increased by

3 seconds

in Feb. 2014.

25

10

20

5

2013

2014

2015

2013

2014

2015

4

5

From Wall St. to Bowling Green

40 seconds

35

Travel time increased by

12 seconds after signals were

modified in March 2013.

30

25

20

2013

2014

2015

4

5

From Fulton St. to Wall St.

25 seconds

20

15

Increased by

3 seconds

in Feb. 2014.

10

5

2013

2014

2015

Source: Internal M.T.A. documents | Chart shows the median travel time between stations.

The analysis stated that if the M.T.A. had known the signal changes would reduce the number of trains able to run on congested lines, they would not have been made.

But the damage was done. After the signal changes, two fewer trains could run on the southbound 4 and 5 lines hourly, forcing the thousands of passengers those trains would have carried to squeeze into already crowded cars.

Across the entire system, more than 1,800 signals have been modified since 1995.

New safety rules meant to protect workers have led to more delays.

After two track workers died within a week of each other in 2007, the M.T.A. created a track safety task force.

The task force recommended new slow zones adjacent to tracks where crews were working; an increase in the minimum crew size; and a longer, more careful setup process for work crews.

The slow zones, marked by lights and flaggers to alert train operators, contributed to an increase in work-related delays.

With the new rules, trains traveling near track work must go at less than 10 m.p.h. — 30 m.p.h. slower than the systemwide speed limit. Even far from track work, the slow zones create bottlenecks and reduce the number of trains able to run.

Delays caused by track work increased each time new rules were added.

10 trains delayed per incident

Slow zones

lengthened

Service stopped during dangerous setups

5

New task force rules implemented, including slow zones

0

2000

2005

2010

2015

Delays caused by track work increased each time new rules were added.

10 trains delayed

Slow zones

lengthened

Service stopped during dangerous setups

5

New task force rules implemented, including slow zones

0

2000

2005

2010

2015

Delays caused by track work increased each time new rules were added.

10 trains delayed

Slow zones

lengthened

Service stopped during dangerous setups

5

New task force rules implemented, including slow zones

0

2000

2005

2010

2015

Source: Internal M.T.A. documents | Incidents are a period of planned or unplanned weekday track work.

As the M.T.A. adopted more safety rules, the share of overall delays attributed to planned track work increased from 20 percent in 2010 to 30 percent in 2014, despite a similar amount of work each year.

Protecting workers is an important part of the M.T.A.’s mission, but the tracks are still dangerous after these new rules. In the last five years, three more workers have died on the tracks, and near misses are not uncommon.

The London Underground, a system of similar size and age, has had no track worker fatalities since 1998.

Tony Utano, the president of the Transport Workers Union Local 100, said he would fight any rule change that could put workers at risk.

“Many of these protections were put into place because our members were maimed and killed on the tracks,” he said.

In an email, Mr. Weinstein of the M.T.A. said that “the safety of workers and riders is our Number 1 priority,” declining to answer specific questions about the increase in delays because of track work.

New York City’s subway is the world’s largest 24/7 system, making it harder to protect track workers while doing maintenance without disrupting service. And even with Fastrack, when some tracks are shut down overnight to allow for a safer and faster job, the start of work can be delayed for hours.

For years, overcrowding was incorrectly blamed for being the root cause of delays.

As delays began to rise in 2013, the M.T.A. blamed rising ridership and overcrowding. But delays kept increasing even after ridership plateaued and fell.

Train dispatchers have often used overcrowding as a catchall category for delays without a clear cause or for ones caused by the crowded platforms that follow when equipment failures or track work disrupt regular service.

Size: 720

Almost all of the increase in delays was in the catchall “overcrowding” category ...

250 incidents per weekday

200

Overcrowding or unknown

150

100

Other delays

50

0

2012

2013

2014

2015

2016

2017

2018

... as subway ridership stayed nearly flat.

6 million

4

Weekday riders

2

0

2012

2013

2014

2015

2016

2017

2018

Almost all of the increase in delays was in the

catchall “overcrowding” category ...

250 incidents per weekday

200

Overcrowding

or unknown

150

100

Other delays

50

0

2012

2013

2014

2015

2016

2017

2018

... as subway ridership stayed nearly flat.

6 million

4

Weekday riders

2

0

2012

2013

2014

2015

2016

2017

2018

Almost all of the increase in delays was in the catchall “overcrowding” category ...

250 incidents per weekday

200

Overcrowding

or unknown

150

100

Other delays

50

0

2012

2013

2014

2015

2016

2017

2018

... as subway ridership stayed nearly flat.

6 million

3

Weekday riders

0

2012

2013

2014

2015

2016

2017

2018

Source: M.T.A. ridership and delay data

Average weekday ridership actually decreased by 40,000 people from January 2013 to January 2018, while trains went from being on time 84 percent of the time to 58 percent.

The M.T.A. recently acknowledged overcrowding was not actually at the root of the problem. The authority has also blamed old equipment, but train dispatchers have not recorded more instances of electrical or track problems causing delays despite maintenance cuts.

Seemingly small changes have added up to a major impact on the subway’s resilience.

To understand how these slowdowns led to the current crisis, we can simulate a line of subway trains moving as minutes pass. Each car here represents an entire train.

Here are trains on a line with no delays, each one neatly following the other.

To see the pattern of trains as they complete trips, we use diagonal lines to trace out how far each train has traveled over time.
Here’s what happens when a train is delayed: As minutes pass, it moves downward in time, but does not continue forward on the line until the delay is over.
Now let’s add a more realistic safe distance between trains. We can see how one delayed train forces those behind it to wait for “train traffic ahead.”
Let’s zoom out and tally the number of trains that are able to finish the trip over 60 minutes.
If we look at how the subway operated before the signal changes, we can see that even when a line had delays, the system had extra capacity to recover.

By moving this slider you can try increasing the number of delays to see what happens to the trains.
We can simulate the intended impact of the M.T.A.’s signal changes. Try moving the slider to increase the spacing between trains.

Up to a point, extra capacity can be removed from the system to increase safety margins without decreasing the number of trains that can run.
Because the new signals are often faulty or poorly maintained, they end up increasing the spacing between trains more than anticipated.

The unpublished M.T.A. analysis from 2014 found that the new signals had significant effects.
To simulate the impact of additional safety rules during track work, we can increase the number of delays again.

Removing slack has an even bigger impact when there are more problems than normal. Delays cascade down the line, trips take longer as trains wait for traffic ahead of them to clear, and the system struggles to recover.

The subway has been operating at the edge of its capacity for disruptions, so small increases in delays and spacing have a disproportionate impact on the number of trains that can move through the system.

This is why small changes have slowed the subway system so much in recent years.

This won’t be easy to fix.

Before the 1995 Williamsburg Bridge crash, the M.T.A. had installed less expensive brakes with longer stopping distances without adjusting the signals to compensate. Afterward, the M.T.A. not only increased the distance between trains but also reduced the speed of train cars and installed speed-limiting signals.

Either lowering speeds or increasing spacing would have given the train enough stopping distance to avoid the accident. Since then, making both of those changes has brought the system to its current crisis.

When asked directly about the adjustments, Mr. Weinstein said repairs and upgrades were continuing as part of the Subway Action Plan, the M.T.A.’s $836 million emergency effort to improve the system.

Just 7 percent of the Subway Action Plan budget had been set aside for signal maintenance as of March 1.

When asked in September how long it would take for riders to see the impact of the Subway Action Plan, Gov. Andrew M. Cuomo told reporters, “I would venture to say if you were looking very carefully, you would see improvement already.”

But the subway kept slowing down. As of February, the number of delayed trains was up about 8 percent since last September.

Note: The train simulation is simplified for clarity. It leaves out important factors that would make delays spread even more. When trains are late, crowds form on platforms and slow service further. Late trains do not make their scheduled merges with other lines, spreading delays across the system.