The Paper Schedule

There’s a paper on the wall of every train station in Tokyo. It tells you when the next train will come. Not “in 4 minutes” or “delayed” or “cancelled” on a digital display that updates whenever it feels like it. An actual, printed schedule, with actual times, down to the minute.

Best part. The trains come when the paper says they will.

I moved to New York for school about 2 years ago. I take the subway every day. The first time I came home from Tokyo and stood on a platform waiting for the Q, I started thinking about something I’d had wrong for a long time.

For most of my life I assumed the difference between New York’s subway and Tokyo’s was money. That if the MTA was just willing to spend more, the trains would work the same way. I really hate being wrong about this, but I was.

The MTA has a $21.3 billion operating budget^[1]. Per rider, it is one of the most-funded transit agencies on the planet. New York pays for its subways many times over, whether in fares, dedicated taxes, federal grants, or the congestion pricing fight that took most of a decade to settle. Only 39%^[1] of that budget comes from fares and tolls. The rest comes from taxes and government contributions, meaning the entire region has been quietly subsidizing the system for years. The money is there. The problem is what happens after we hand it over.

I’ve spent some time looking at the MTA up close. I built a fare analysis^[2] last year showing that New York’s flat $2.90 fare (now $3) could be replaced with a distance-based model that would generate around $913 million in additional annual revenue while lowering costs for about 15% of trips. The deeper I went into the data, the clearer it became that the system’s problems aren’t hiding. The numbers are fine. They have always been fine.

What’s actually broken doesn’t show up on a balance sheet. It shows up in who answers for the trains when they don’t run. It shows up in how much it costs us to build a single new mile of track.

Who actually answers when the trains don’t run?

The answer? It depends, and that’s the problem.

A 17-member board governs the MTA, and every single one of them is technically appointed by the Governor of New York. The state Senate confirms them. Of those 17 members, 6 are nominated directly by the Governor, including the Chair, who runs the entire agency. Four more are nominated by the Governor on the recommendation of the Mayor of New York City. Three are recommended by the county executives of Nassau, Suffolk, and Westchester. The remaining four are recommended by executives from Dutchess, Orange, Putnam, and Rockland counties, and together they cast a single collective vote^[9].

NYC Transit handles roughly 2.4 billion rides a year. Metro-North and the Long Island Rail Road combined handle about 175 million. Add those up. 2.575 billion total rides. NYC Transit’s share is 93%. So if you ride the MTA, there’s a 93% chance you’re riding inside New York City. And of the 17 board seats, only 4 are filled by people the Mayor of New York City had any say in. That’s 24% of the board. 93% of the riders, 24% of the board.

The actual people getting on the subway in the morning have no one who directly answers to them.

This is what governance looks like in practice.

In late 2018, Governor Andrew Cuomo personally took control of an MTA decision^[10]. The agency had spent years preparing for a 15-month shutdown of the L train so it could repair the Canarsie Tunnel, which had been flooded with salt water during Hurricane Sandy in 2012. The plan had gone through extensive public hearings. Businesses had relocated. Commuters had built workarounds. The MTA’s paid consultants (WSP and Jacobs Engineering, on contracts worth hundreds of millions) had explicitly endorsed the full shutdown as the “preferred approach”^[11].

Then, three months before the shutdown was supposed to begin, Cuomo called the engineering deans of Cornell and Columbia and asked them to take a look at the plan. They worked unpaid over winter break. Three weeks later, they came back with a different proposal: wrap the corroding bench wall in fiberglass instead of rebuilding it, and run new cables along the tunnel walls instead of encasing them in concrete. On January 3, 2019, Cuomo held a press conference announcing the shutdown was canceled. The MTA chairman stood next to him and endorsed the new plan. No MTA officials publicly raised any objections. The shutdown never happened. The new plan worked.

The detail that haunts me about this story is what was on the MTA’s payroll while it was happening. 173 civil engineers. 107 project managers. 73 architects^[11]. None of them had proposed the academic plan. The hundreds of millions in consultants couldn’t see what two unpaid deans saw in three weeks. The Riders Alliance called it “a last-minute Hail Mary idea cooked up over Christmas,” and they meant it as an insult, but it stuck precisely because it was true.

You can read the L train story two ways at once, and both are correct. One reading is that the MTA is so dysfunctional that years of expert planning could be improved by free academic labor over a holiday break. The other reading is that one elected politician unilaterally overruled years of public process with a plan invented in three weeks by people accountable to no one. Whether or not it worked this time, the next time it might not. There is no functional system between the people running the trains and the people riding them.

This is the structural answer to who is in charge of the subway. The Governor is in charge. The Governor has always been in charge. The Governor is elected by people who, for the most part, don’t ride. The people who do ride get to elect a Mayor who recommends four members to a board controlled by the Governor. The accountability gap is not an accident. It was built into the agency at its founding.

Now the cost.

The accountability gap doesn’t stay in the boardroom. It shows up in what the agency builds, and what it pays to build it.

A single mile of new subway track in New York costs about $2.5 billion^[3]. Phase 2 of the Second Avenue Subway, still waiting for the rest of its funding, is projected at around $4.3 billion per mile^[4]. The global average across 900 projects in 60 countries is closer to $350 million per mile. Similar projects in Italy, Sweden, Spain, Turkey, France, and Germany typically come in between $250 and $450 million. Seoul builds at roughly one-twentieth^[5] of our per-mile cost.

$2.5 billion divided by $350 million is 7.1x. That’s Phase 1 of the Second Avenue Subway versus the global average. Phase 2 at $4.3 billion per mile is 12.3x. And Seoul builds at one-twentieth our cost, meaning Seoul builds a subway for roughly 5% of what we pay. We are not 20 times worse at moving dirt or pouring concrete. We have just become 20 times worse at running the operation that does it.

There are limitations to this comparison, and I want to name them before someone else does. Tunneling in Manhattan is genuinely harder than in some cities. Older utility infrastructure complicates everything. American labor and material costs run higher in general. But the Transit Costs Project, which has spent the better part of a decade on exactly this question, controls for most of those. The 20x gap doesn’t disappear when you do.

The Transit Costs Project^[6] at NYU’s Marron Institute points to 3 clear pain points. Physical structures. Labor. Procurement.

I’ll start with the stations.

The Second Avenue Subway’s mined stations are roughly twice as long as the trains that run through them. The 96th Street Q station is closer to three times longer. By comparison, Italian and French subway stations are 5–10% longer than their trains. Stockholm’s Odenplan is 17%. Stand on the 72nd Street platform at some point. The station feels luxurious and grand. High ceilings, mosaics, a vast tunnel-like feel, and the steepest escalators known to man, all of which beat most of the rest of the system. But the train pulls in and there’s more empty platform than there is platform for trains to operate on. By no means am I saying the station isn’t beautiful. It’s just way larger than it needs to be. Phase 1 of the Second Avenue Subway used two different escalator vendors across three stations^[7], with each station essentially custom-built, even when standardization would have been cheaper and faster.

Next, labor. In Turkey, Italy, and Sweden, labor costs are roughly 20–30% of total construction^[6]. In the Northeast US, labor is about 40–60%. That difference doesn’t come from American workers being paid more. At most a quarter of the gap is wages. The rest is headcount.

During Phase 1 of the Second Avenue Subway, the tunnel boring machine was operated by 46 workers per 8-hour shift, with another 29 supervisors overseeing them across the day. The Transit Costs Project’s interviewees said the same machine could have been run with 30 workers per shift, and that the number of supervisors exceeded any international norm.

Across the day, 46 plus 29 is 75. 75 divided by 30 is 2.5. We are sending 2.5 times the labor required to do the work.

Finally, procurement. Different agencies that share a worksite (utilities, transit operators, the city, the state) each insist on sending their own supervisor into the tunnel to watch the contractor. Each set of supervisors gets paid out of the project budget. The Transit Costs Project describes the broader pattern as “a pervasive culture of secrecy and adversarialism between agencies and contractors,” combined with a lack of capacity at the MTA to manage the consultants it hires. The result is a system in which paying for oversight has become a major cost driver, and the oversight itself doesn’t seem to make the work any better.

It is important to me that I tell you what this argument is and what it isn’t. It isn’t an argument that MTA workers are overpaid. The research is explicit that pay differences account for at most a quarter of the gap^[7]. It isn’t an argument that unions are the problem. White-collar overstaffing is just as severe. And it’s not an argument that New York has caught some uniquely American disease that can’t be cured. Alon Levy, one of the Transit Costs Project’s lead researchers, has given it to us straight^[8]. “The causes of high American (especially New York) costs are institutional, and fixable, without a revolutionary upending of any legal or social system.”

The important piece is that we aren’t fixing them. In late 2024, the MTA approved roughly $250 million in new consulting contracts for Phase 2 of the Second Avenue Subway. The same project the Transit Costs Project used as its case study for procurement dysfunction. The pattern the researchers identified as a primary cost driver is being repeated, on the same line, in plain sight.

None of what I’ve laid out is mysterious. The Transit Costs Project has spent years documenting it. Reporters have spent decades on it. The MTA itself publishes most of the relevant numbers in its annual reports. The diagnosis isn’t hidden. The governance structure that lets a governor override years of expert planning is written into the public authorities law. The construction costs are public record. The procurement dysfunction is named in a published academic report whose authors will happily give you a tour.

What we don’t have is a political incentive to fix any of it. The governor isn’t punished by NYC riders for bad decisions, because most of his voters don’t ride. The MTA isn’t punished by the public, because the public can’t tell whether their delay is the agency’s fault or someone else’s. And the consultants keep getting hired because the people doing the hiring don’t have the internal capacity to do the work themselves. Which is the same problem the consultants were hired to solve.

I think about the paper schedule in Tokyo a lot. Not because the United States is going to become Japan, but because the paper schedule is evidence that a transit system can be built with the goal of working, and succeed at it. We do not need to copy Tokyo. We need to want this for ourselves.