Why Are BART and MUNI always broke(n)?

If we want to fix our transit, we have to understand why they fail.

A MUNI train, circa 2022

Once again, San Francisco’s main public transit systems—the Municipal Railway (MUNI) and the Bay Area Rapid Transit (BART)—are broke. MUNI’s annual budget is ~$1.49 billion, but it has a $307 million deficit going into 2027 (slated to expand to $434 million within five years). BART’s annual budget is ~$2.3 billion, and it faces a $376 million shortfall, which is also posed to expand in 2028.

Once again, San Francisco voters are asked to intervene, with two upcoming measures: an additional property tax on all real estate in the city, and a supplemental 1% sales tax on all transactions in the city. If both taxes pass, the median household in San Francisco will see their tax bill increase by $450.

And if either tax fails, both BART and MUNI will make absolutely massive service cuts. For MUNI, the long-tail of lesser-used bus lines will be eliminated. All “rapid” bus lines will be canceled. Frequencies will be cut by 50% across the board. The underground section of the MUNI will stop running after 9PM. For BART, there will be 70% frequency cuts across all lines. All trains will stop running after 9PM—including the airport lines. Fifteen stations will close. And parking fees will (at least) increase by 50%.

This is the apocalypse scenario for public transit. Service cuts will destroy ridership, which will have a major impact on revenue. That will also lead to lower state subsidization, which will worsen service, which will itself reduce ridership. This locks the system into a negative feedback cycle which will compound into a generational disaster for transit in San Francisco.

If these taxes fail, the system will not recover for a decade.

Given the scale of our impending disaster, there’s a seemingly obvious answer to the tax question: vote for them. The most technologically important city in the world, where ambitious acolytes labor to conjure a digital God, should have trains that run after 9PM.

This is the angle advocates have taken. The system is broke, transit is important. “A fully functioning Muni is essential for that bustling economy,” they say. And they’re emphasizing the doom. “BART is not too big to fail.” They mention the second-order effects too. SF’s mild post-pandemic recovery will be extinguished and the city will return to its 2020-era gloom. House values in outlying suburbs will crater. Job creation will decline and labor shortages will multiply.

I don’t disagree with any of these impacts. Assuming these agencies aren’t inflating the impact, and the crisis does seem dire enough that they aren’t, the advocates are right. Vote yes, because while taxes suck, the downside of a broken transit system is unbounded.

But there is a certain moral exhaustion this rhetoric brings. Living in SF, it seems like every public service is perpetually on the brink of crisis—always one additional tax from utter bankruptcy—and there’s always some reason to pay more. The appeals for money, each time, always read like moral exhortations tinged with noblesse oblige, where you should pay because you can pay, and you shouldn’t question much more than that. Efficiency, effective usage, cost controls: these barely even enter the discussion.

To me, this begs the obvious question: is our transit actually underfunded? Or does it just spend its money poorly?

The DOGE Angle

The most common explanation I hear about BART and MUNI’s eternal decrepitude are its wasteful labor costs. In an odd similarity, we hear this about colleges, about public schools, about healthcare, and now, about transit, and you hear it ever more often now because of DOGE. Elon claimed to uncover trillions in waste in government; even now, nearly a year after his departure, he spent thirty minutes on the Dwarkesh podcast railing against long-dead social security recipients dutifully collecting their fraudulent billions.

At the federal level, his claims have dubious veracity. But MUNI and BART do have considerable spending on labor. Perhaps there’s something different here.

In MUNI’s operating budget in FY2025, about $919.1M—nearly two-thirds—goes to labor. Another $271.4M share goes to “professional services,” which includes spending on consultants, software systems, legal services, and IT support. The final ~10% goes to materials and debt service. For BART, the breakdown is similar. Out of a $1.2B operating budget, $763.3M—about two-thirds—goes to labor, $243.4M is for non-labor expenses (including parts, maintenance, and the “professional-services” category above), and another $145M goes to debt service [7].

That does seem like suspiciously high labor spending.

What does labor refer to here? SFMTA, in a section designed to respond almost exactly to this line of questioning, states that “80% of labor costs go to operators or direct support, and 97% of employees are non-management.” In aggregate, there are 6,292 budgeted FTE, with 5,169 in transit. Out of those 2,670 are transit operators (bus drivers and train operators), 1,003 in vehicle maintenance, 676 in line management, 445 in infrastructure maintenance, 348 in cleanliness and security, and 27 in safety compliance. Out of remaining ~1000, 601 are in revenue generation, a broad category including parking citation officers, fare inspectors, meter operations, and parking facility staff, 350 are in admin, 142 are in street management and safety (crossing guards, signal technicians, etc) and 35 are in various minor transit initiatives.

BART shows a similar background. They have 3,760 FTEs. Out of these, ~1000 are train operators, station agents, and other train-related employees, 763 in train and track maintenance, and the remaining in safety and compliance. Another 409 are BART police, hired to enforce order. The remainder are administration staff: HR, finance, IT, land-use and planning, infrastructure development, communications, and our favorite bugbears: civil rights and DEI.

From these breakdowns, I don’t see many obvious cuts. MUNI’s administration seems fairly small, and while BART’s is considerably larger, the service itself covers much more ground. There is probably bloat, but not enough to actually dent the budget. Yes, the DEI section is indeed quite useless, but cutting it saves almost nothing.

Why do we need so many transit employees? The short answer is that running trains is a labor-intensive business with minimal automation, which is worsened by strict union requirements.

For the long answer, let’s start with BART. The Amalgamated Transit Union (ATU) allows for 12 hour shifts per operator, with 11 hours of mandated rest. Part-time operators are capped at 25 hours a week and 15% of the workforce. BART itself says it only runs 62 trains simultaneously at peak hours. If you consider each running train a “slot,” then to cover that slot, you need at least two operators a day. You actually need more, because not every operator does a 12 hour shift each working day, and if they work more than 40 hours a week they get overtime. You also cannot keep operators “around” for longer than 12 hours: that is, even if peak hour is passed and the train count reduced, you can’t make an operator “off-the-clock” and reactivate them during the next peak segment—you’re billed for the continuous segment and you need new operators after that 12 hrs. Putting that all together, you need at least 4-5 operators per slot, without accounting for vacations, weekends, sick days, trainings, and “wow I really need to pee right now so let’s swap” emergencies.

Sure, the union requirements are quite rigorous. The biggest issues are the restrictions against hiring part-time employees and banning split-shifts, but even those would probably help more in service flexibility rather than enabling widespread cuts. And there’s also the rest of the staff. There are people to coordinate the network, handling track changes or service disruptions, people to maintain not just the trains (which require constant maintenance) but the actual tracks and the electrical delivery and the transponders, people to clean each train, people at the stations to handle tourists who never learned how to read, and so on. This adds up to a lot of manpower. According to my crude estimate, every train “slot” has something like 5 secondary support staff required.

For Muni, the same logic applies with a larger denominator. I will spare you the exhaustive detail and say only that their union contracts are very similar: 12 hours maximum without rest, 12 hour split-shift maximums, and 40 hours a week before overtime. The difference is the MUNI needs even more operators, because instead of 62 simultaneous trains, you have, at peak, 364 buses, 158 trolleys, 109 light-rail vehicles, 14 streetcars, and 26 cable cars (data from 2024). This isn’t one-to-one, as 2-3 light-rail vehicles are connected into a single train with a single operator, so let’s make a low-end estimate of three vehicles per train and say 36 trains. Buses, light-rail, and street-car take one operator each, cable-cars take two. This means we have ~570 simultaneous slots, for which MUNI has 2670 operators. Similar ratio as BART.

Who Needs The Operator?

Okay, you then ask—or at least I do—can’t we automate this? Driverless trains are a solved problem: I went to Vancouver two years ago and they worked great. And the answer is yes we can, yes the unions are opposed, and—it might make a difference, but not enough, and not soon enough.

First, BART trains are already largely autonomous. The train starts, stops, and changes tracks on its own, without operator intervention, since at least 2024. This is actually pretty easy, because trains don’t need to steer, don’t have to interact with other traffic, and have predictable acceleration and deceleration patterns. Really, it’s closer to an auto-throttle than any form of road autonomy.

There are many things this auto-throttle can’t handle, however. The biggest issues are track obstacles and platform safety. Vancouver’s SkyTrain has LIDAR and infrared lasers installed throughout the system to detect track hazards. BART doesn’t have this, and it would take considerable expense to install. The other dimension is, the platform after the train has stopped is the most dangerous part of transit. What if a pedestrian falls onto the track? What if their jacket gets caught in the door as they step out? The way fully autonomous systems handle this is with platform screen doors (PSD) and sensors. This is hard for BART for a few reasons. First, this requires trains to stop on the same spot each time, which BART can’t do, as it hasn’t fully modernized its train-tracking system yet. BART’s CBTC modernization is still ongoing, and slated for 2030. Second, PSD is very expensive, in part due to BART-specific engineering constraints, and in part because capital projects in American transit are extremely expensive for various reasons that we will cover another day. And last, BART already has issues with vandalism and antisocial behavior. What happens when someone smashes the glass on an automated door and bricks the entire station until it can be repaired?

Okay, but let’s assume all of these problems have already been solved and we can fire every operator right now. BART doesn’t publish the number of operators it has, but to make a large overestimate, we can add up the rail department sizes (which include many non-operators), giving us 385 employees. From what I see, the highest reported salary for a train operator in 2024 was 209,000, with more datapoints in the 100k to 150k range. If we assume 150k for each, which is high, then cutting all of them saves us $58 million. A lot, but not nearly enough.

None of this is to say that autonomous trains are undesirable and shouldn’t be done. If all of these modernization efforts are seen through, we’ll have a system with higher peak capacity, higher frequencies, and the marginal cost of expansion will decrease considerably. It would be a huge win. But is not anything close to a short-term salve.

For MUNI, full autonomy is a much harder problem. Where BART is a grade-separated railroad for its entire length, MUNI Metro trains spend much of their time above-ground, sharing normal roads with cars and pedestrians. This makes the base autonomy problem much harder than that of a train. Obviously, the same goes for buses, which, apart from a few special cases (Geary, Van Ness, Mission), run in standard lanes with normal street traffic.

For MUNI Metro, there are definitely opportunities for increased autonomy. The underground segment of the metro is the easiest target. Like BART, the modern trains have an auto-throttle mechanism to make normal operation autonomous. Further autonomy there has the same issues: platform doors, which are also expensive, and an in-progress train control upgrade to allow for higher capacity and more precise stops—the current one still uses floppy-disks. Still, we can imagine a future where only one line runs in the tunnel, fully autonomously, and passengers just walk upstairs to transfer to above-ground lines. This would enable some labor savings.

Unlike BART, it’s quite hard to determine how many operators we could shed here, as you still need them above-ground. Why? Because above-ground, the tech needed is essentially that of a self-driving car. But, you say, we have already self-driving cars on the streets of San Francisco. Why can’t we just license their technology?

And we can. This probably is the future. Over the next two decades, fleet modernization will introduce self-driving buses and trains to the system and they will improve service significantly. But there are a multitude of issues that need to be resolved first. The environment of a self-driving car is quite different than what a San Francisco bus or aboveground train runs in. A bus or train stops constantly, at specific spots, where they facilitate mass onboarding and offboarding. Every stop has the traditional hazards we discussed for trains, like obstacles, people stuck in the doors, along with new issues, like wheelchairs, the “kneeling” system of the bus, and obstructions at the transit stop. SF’s dysfunctional social environment also comes into this. MUNI operators help prevent aggressive people from boarding, and stop antisocial behavior on the bus itself (like drug use). And there’s also the sheer cost of it all. We need a new generation of vehicles with a network of sensors and software. These will take years to commission, approve, and acquire.

The Scrooge McDuck Story

Okay, if we can’t just fire all of the employees, can we pay them less? BART and MUNI are rife with overpaid public employees suckling on the taxpayer’s teat. There’s a janitor making $270k a year (while napping in a closet), a BART escalator mechanic making $210k, and train operators making 285k. And even worse, BART and MUNI are hiring. Between 2019 and 2024, BART’s headcount went from 3,985 to 4,292, while MUNIs went from 5,981 to 6,798. Despite their growing deficits, BART and MUNI grew, adding new useless overpaid positions to their bloated parasitical administrations. We can obviously just cut salaries across the board and find massive savings.

Except we can’t, at all. The examples above have bad optics but are not typical. The average transit operator makes less than 150k, and “System Service Workers,” which is the BART title for janitor, average ~80k, with the highest salary reported at $176,221. The janitor making 270k apparently worked 361 days a year, taking every available shift, including weekends and holidays. At his highest stretch, he worked 17 hours a day for 18 straight days.

And while the average pay may seem high for a sanitation employee, consider the job more broadly. BART stations are filthy and constantly reek of piss and shit. Who cleans that? BART reopened its bathrooms recently, adding a janky self-washing mechanism that probably doesn’t work. Who do you think is the backstop there? This man worked almost every day of the year and he still ended up making less than the do-nothings that build JIRA. It seems fairly reasonable to me.

Every “egregious” example of a highly paid transit employee involves massive overtime shifts. Why do these overtime shifts exist? Because the agency is understaffed for the amount of work it has to do, and in almost every case, it is cheaper to have an employee do overtime than to hire another employee to split the work. There have been issues around overtime spend tracking, but this is more about bureaucratic attribution rather than widespread fraud, and really just underlines the core point.

On average, transit employees make considerably less than those headline figures. As we’ve already discussed, the average train operator makes ~120k. The average janitor makes ~85k. The higher average wages are for tradespeople, but often, these people are taking a pay-cut relative to their market rate. HVAC, electricians, escalator techs, elevator techs, and the like are all in short supply in the Bay Area. They’d make more in private business, but they work at BART/MUNI for pensions and benefits.

Capital Expenditures

In the last few sections, we’ve discussed the operational budget for BART and MUNI and found that while there certainly are economies to be had, they aren’t sufficient to cover the deficit. The real efficiency gains on offer actually require more short-term investment to drive greater efficiencies over the next few decades.

But operational budgets are only half(ish) of the story. BART has a $1.2B operational budget, but a $1.1B capital budget. MUNI has $1.45B operational budget, and a $423M capital budget. If there is a budgetary crisis, can’t each system use the capital budget for operational expenses, instead of damaging service cuts?

Like everything in this essay, the answer is maybe, with deep caveats. To start, MUNI actually was doing this already during the early stages of the pandemic. Even before the pandemic, it was already facing an operational deficit. They had a flexible pot of money from the city’s general fund that they repurposed from maintenance and vehicle replacements to use for operational costs. By 2022, this ran out, and the city was planning to cut the meter modernization program for the budget, but an injection of federal money via the CARES Act (signed by Donald Trump, ironically), backfilled the capital budget and saved the project.

In 2022, the bipartisan infrastructure bill created Section 5307 funds, which disbursed funds to local transit systems. These were targeted for capital expenses, but the law allowed for preventative maintenance and paratransit, which fall under MUNI’s operational budget. Thus, these were another backstop.

BART, on the other hand, didn’t have a similar mechanic. Before the pandemic, it was a sustainable system running under-budget, and actually shifted operational expenses to capital expenditure. This was useful to capture federal programs that matched local spending, allowing much larger capital projects than would be otherwise possible. The system underwent considerable expansion over the past decade, adding several new stations between Fremont and San Jose, and replacing the last generation of train cars with modern ones. It was never able to use capital expenditure for operational expenses.

Now, both systems have run dry on flexible capital funds. The remainder of the capital budget comes from bond measures, voted on by the public, which are approved for only a specific purpose. A large part of BART’s budget comes from Measure RR, a $3.5B bond measure approved by the public in 2016 to fund critical track and tunnel upgrades and capacity improvements. MUNI’s budget comes from various bond measures since 2016 with the same restrictions. Ironically, successful bond measures that increase the capital budget actually increase the operational deficit, as debt service has to be an operational expense.

To preempt the next question—no, we should not use bond measures for operational expenses. This is a basic good-governance measure and the intuition for it is clear. Passing new taxes or raising fares is difficult because it makes costs visible to voters. Bonds, on the other hand, are indirect taxes that are easy to obscure from voters. If politicians can just issue new debt each year, there’s never an incentive to economize or build reliable funding mechanisms. It will lead to increasingly large debts, spiraling debt service, and eventually, system collapse. This is the same mechanism as countries (that aren’t the US) issuing long-term debt for short-term expenses: it will always lead to a credit-crunch.

What Caused the Crisis?

So far, we’ve proven that MUNI and BART aren’t outrageously wasteful, haven’t massively overhired, aren’t massively overpaying janitors, and don’t have a hidden pot of money to solve their operational deficit. But we haven’t discussed why the deficit exists in the first place.

The first dimension here is revenue collapse. For BART, the explanation is simple—and especially tragic. Before the pandemic, BART was the second most revenue-efficient transit system in the country (after its Bay Area counterpart, Caltrain). In 2019, BART fares and parking generated $520M, which is about 66% of its operating budget for that year. The rest was covered by sales taxes and property taxes across the five bay area counties. In 2021, with lockdowns and work-from-home, fare and parking revenue fell to $69M, a 650% decrease. No organization can survive that degree of revenue collapse. BART, as as system, is centered around commuters heading to downtown SF from the peninsula and from the east bay. After the pandemic, the Bay never returned entirely to an in-person economy, and so, BART revenues have never fully recovered. In 2025, they were only up to $259.1M in fares and parking revenue—less than half of where they started.

This may sound dramatic, but think about the numbers. Very few office-workers are doing full 5-day RTO. If they’re doing 4-day RTO, then that’s a 20% decline in revenue already. Most people I know are somewhere between 2 and 3, varying based on the week, and they often have other dimensions of flexibility. For example, maybe now they commute to a local office rather than the HQ in the city, or they come in later after traffic subsides and then can drive for the convenience. These are all widely accepted work patterns now, but before the pandemic, each of these was a BART commute.

MUNI has a much more mixed story. It had a growing operational deficit well before the pandemic started. It has never covered much of its operations through fares—in 2023, the farebox recovery ratio (the ratio of fares to operational expenses) was just 8.6%, and even pre-pandemic it peaked at 24.5%. It’s revenue comes from citywide parking fees, various state-level programs for transit funding, large disbursements from the SF General Fund, and, since 2022, federal relief (that we discussed earlier). The latter is running out, leading to this deficit.

But this isn’t the entire issue. These deficits are in part due to an exogenous shock that fractured their funding models and left them in the lurch. This, however, neglects various structural issues that already left both systems especially vulnerable to ridership declines.

First, MUNI has had an increasing deficit for over a decade, well before the pandemic. The major cause here is the decline in parking revenue, which has been falling each year since 2015. This is due to rideshare companies. Rideshare reduces the number of transit trips, which hurts revenue directly, and it reduces driving, which hurts parking fees. Parking is designed as a revenue hedge against ridership, as it raises transit revenue even when transit can’t service a particular trip, and the decline of this counter-cyclical funding mechanism hurt the system’s viability.

MUNI’s other main vulnerability comes from its dependency on unstructured transfers from San Francisco’s tax base. This isn’t a priori unusual—Toronto’s metro system uses the same mechanism. But SF’s budget is unique in that despite its very large headline number, the actual money that the local government can freely allocate each year is relatively low.

In 2026, San Francisco’s total budget was ~$15.95 billion. Of this, some ~9B is pre-allocated. About ~$2.1B went to the SF Public Utilities Commission, $1.8B went to the airport, $1.4B went to MUNI (not true in 2027!), and $156M went to the Port of SF. Another ~$500M went to overall debt service. The remaining $7B is part of the general fund, but about $2.7B of that is already mandated from various historical ballot propositions and laws.

Remove certain classes of revenue, and that leaves $4.6B, which is what the city calls Aggregate Discretionary Revenue (ADR). Out of this, the city funds the police ($1.9B), the hospitals ($2B), general city administration ($150M), and then, the long-tail of everything else a city has to do: public defenders, prosecutors, the DA, the fire department, elections, and so on. This is rarely sufficient: the city already makes cuts somewhere every single year.

And so, when MUNI’s federal funding expired, despite the city’s on-paper abundance, almost every dollar was already spoken for. MUNI’s dedicated revenue streams had been weakening, its outside funding injections were drying up, and the city it’s servicing doesn’t have any money to spare.

BART’s vulnerabilities are almost the opposite of MUNIs. Where MUNI never relied significantly on fares, BART’s solvency hinges on fares. It only runs four lines that are optimized for commute patterns, and it optimizes its frequency around the usual 9 to 5 workday. And so, when COVID destroyed old ridership patterns, BART was left vastly over-provisioned for its initial use-case, but unable to cut costs without cratering its remaining ridership. BART would say its current predicament is a punishment for its success as an efficient commuter rail; I would say instead that the lack of revenue diversification was a known vulnerability that was never addressed because politicians didn’t think it was worth the effort.

But even if BART and MUNI had better funding structures, they’d still face the same incredibly high (and growing) labor costs. The root cause, as far I can deduce, is that transit is an inherently labor-intensive service, and the costs of labor in the Bay Area are very high. There are two main reason for this: housing prices and cost disease.

First, for most people, housing is their largest expense and the core determinant of their cost of living. This means that high housing costs raise the floor for the acceptable wage in any given area, which also drives up the cost of every labor-intensive service present there. This is one of the core tenets of the housing theory of everything, and it means for MUNI (and BART), to retain its employees, has to pay at least enough to match perpetually rising rents in San Francisco. How bad is it? Consider this: in SF, a family of four can earn up to $155,850 and still qualify for affordable housing.

And this leads to the next problem. It’s not enough for MUNI to have salaries that just match the cost of living, because government agencies have to compete with the private sector to hire. This requires roughly matching their salaries. But what’s the private sector in the Bay Area? For white-collar professions, it’s technology, AI, biotech, finance, and law—high-paying professions that have seen massive growth over the past decade. For blue-collar professions, the Bay Area has some of the highest rates for skilled labor in the country. Additionally, private sector salaries (especially in the white-collar professions) rise because increasing productivity gives companies more money to spend. Transit hasn’t experienced any of the same productivity increases, but it still has to offer competitive salaries. That leads to continually increasing labor budgets to even retain their workforce.

Both of these problems arise from the incredible economic success of the Bay Area, and together, they’re a recipe for perpetually rising operational expenses. To mitigate them, we have to both cut housing costs, which is perhaps the biggest problem in American politics, or increase transit productivity, which requires the long-term automation investments we discussed earlier. Neither solution is feasible in 2026.

Finally, given this is San Francisco, we have to touch on fare evasion. It is a problem, it is widespread, it’s an easy target for public opprobrium—and it barely matters for the top-line operational budget. Really, fare evasion is more about safety and quality of life, not revenue. It’s strongly related to general antisocial behavior on transit, which discourages usage and creates stigma.

In that sense, we actually understate the problem by reducing it to budgetary terms. To this point, BART recently spent $90 million on new fare-gates to reduce evasion, and it has actually reported a major decline in problematic behavior on trains since the new gates were installed. Why? The people that are unwilling to spend < $5 for a train ride have a lot of overlap with the people blasting music, urinating, or harassing women on the train, and so, reducing one greatly reduces the other. The solution for MUNI is quite a bit harder (how do you install gates on buses?), but it any effective solution will likely have the same benefits.

But Isn’t This Also True In <city_name>?

The pandemic wasn’t, of course, localized to San Francisco. Every single city in the world grappled with lockdowns, the sudden absence of foot traffic in downtowns, and the radical alteration of commute patterns. Every public transit system had tumbleweeds rolling through their empty trains in April 2020.

Indeed, BART and MUNI certainly aren’t unique for their incomplete recovery. The New York subway had 1.195 billion rides in 2024, which is a 30% decline from 2019. The Los Angeles Metro was at 83% of pre-pandemic ridership by early 2025. Outside of the US, the Toronto metro recovered roughly to NYC subway levels, with ~75% of normal ridership in 2024, and Vancouver is at about 84% of 2019 traffic. In Europe, Paris is at ~90% of pre-pandemic ridership. London is the only large western city I’ve seen that has made a near-full recovery, with 93% of pre-pandemic ridership and 98% of pre-pandemic total mileage.

Yet, we don’t hear about budgetary collapse in New York, Paris, LA, Toronto, or really, any other major city. Why? There are two main reasons.

The first is funding mechanisms. As we discussed above, BART is uniquely dependent on fares for revenue, while MUNI has been relying on an unstable combination of fares, parking revenue, and municipal largesse. This left both of them massively exposed to a ridership decline. Now, consider New York: only 39% of the Metropolitan Transit Authority’s (MTA) revenue comes from fares, while 55% comes from state and local taxes and 13% comes from tolls. This is much more resilient revenue base. Ridership declines did not lead to tax declines, because COVID did not actually cause an economic downturn. And while toll revenues fell during COVID, they didn’t fall nearly as much, and they rebounded faster than transit usage. The recent congestion tax for Manhattan drivers gives the MTA another counter-cyclical source of funding.

You see the same pattern in Paris. Only 34% of the Île-de-France Mobilités (IDFM)’s revenue comes from fares, while 51% comes from an employer-paid payroll tax and 13% comes from general subsidies. The payroll tax is another counter-cyclical force here, because even as ridership fell, employment didn’t, and so the system’s major funding source remained stable.

And second, it’s about the commute vs. non-commute split. The largest systems we discussed above had widespread usage for both commute and non-commute trips. And the recovery pattern for each type of ride is clear: commute-heavy weekday travel had the worst recovery, but non-commute and weekend travel recovered much better, in some cases surpassing its pre-pandemic numbers. This is clearest in New York. In 2024, weekend subway usage was at 80% of pre-pandemic usage, while weekday was at 68%. The Long Island Rail Raid (LIRR), one of the MTA’s commuter lines, has on and off-peak fares, and so can break down commute vs. non-commute rides more precisely. It stated that while commute ridership fell, non-commute ridership was actually up 16.7% over pre-pandemic numbers.

London and Toronto show the same patterns, though their data is less precise. Transport For London (TfL) showed that while ridership was at 93% of 2019 levels, commute-time ridership was down 16%—which is partially offset by growing non-commute ridership. The Toronto Transit Commission (TTC) showed that “commuters”—people using the system four weekdays each week—were at 50% of March 2020 level, but less-frequent users were at 109%, an actual increase. They also show weekend ridership is up to 91%, a significant recovery.

Out of all of these systems, BART is unusually commute dependent. Pre-pandemic, commutes made up 70% of its usage. In 2024, that number fell to 59%. This change happened due to the precipitous fall in the commute segment, and it suggests that the BART network is more useful for commutes, but less so for recreational or discretionary usage.

MUNI is different. It’s a city-focused transit system, with broad coverage and reasonable density throughout its length. The underground metro section is a single trunk line down the city’s core artery, ferrying people to and from downtown, while the aboveground metro and buses are non-radial connectors throughout the city. In 2024, the underground metro was at 59% of 2019 usage, while the aboveground metro and the bus system was at 81%. Most of the latter category was in buses. This means the most cost-effective part of the system faces declining usage, while the labor-intensive coverage-focused sector, designed to be the loss-leader, is gaining ridership. No functioning business can sustain this type of split, and MUNI, unlike a private operator, can’t cut lower-margin routes to sustain its core service. It’s the status quo or broad-based service loss.

What Can We Do?

None of this is an exoneration of BART or MUNI. The systems are not run by a business standard of efficiency. Any private equity firm would find many areas to improve. The problem is, government services do not have the same degrees of freedom as a private business, and in this constrained environment, the available cost-savings are not sufficient to stave off the deficit’s gaping maw.

The main area where we can find many tractable efficiency improvements are transit capital expenditure. San Francisco, in the last decade, has embarked on two of the most expensive transit expansions in the world: the central subway and the Caltrain downtown extension. These are both very short lines in urban areas, and the costs are absurdly high. For example, the Caltrain expansion is a 2.4 mile extension of the Caltrain into downtown SF, and it is slated to cost 7.6 billion dollars—about as much as the 200 mile “Shandong Section” of the Weifang–Suqian High-Speed Railway in China. That railway broke ground in 2024 and is slated to complete in 2028; the Caltrain extension, which is two orders of magnitude smaller, is not expected to complete until 2035.

This is abysmal performance by every possible metric, but it is a form of decrepitude distinct from what we’re discussing now, and deserves its own investigation.

Instead, given our current state, the only two options we have are to increase demand and create new sources of revenue.

To find new revenue, BART and MUNI have to deploy their capital budgets to create a new type of demand. For MUNI, I imagine this is more bus-rapid transit lanes on major arteries, increasing stop spacing on popular routes, and using their new tracking system to prevent busses from bunching up. This will increase frequencies and coverage, helping make the system a more viable substitute to Uber, Lyft, and recently, Waymo, on spoke-to-spoke trips.

For BART, the answer is less clear. One idea is creating new sub-lines focusing on specific parts of the system. You can imagine a line focused on ferrying people up to the city on weekends, and running late so it’s an Uber substitute at night. Once the San Jose Diridon expansion is complete, you can do the same for the East Bay to San Jose circuit. It probably also means shutting down the low-traffic outlying stations to concentrate service in a lower-revenue environment. Maybe also, BART can invest in shuttles from suburban stations to suburban office parks, which makes it more viable for that genre of trip.

And while sales taxes are an initial attempt at stable government funding, they aren’t sufficient. The current math shows the system will face another deficit in three years as the latest state subsidies expire. I think what the MTA has built in New York, with many different counter-cyclical revenue sources, is an ideal to strive for here. One medium-term fix here is a congestion tax. Rideshare has replaced driving as the primary non-transit mode of travel, and so, a congestion tax should replace parking as a counter-cyclical revenue hedge for MUNI. This also prices in the externality of traffic in SF’s downtown core, which will increase transit usage.

More broadly, the Bay Area needs to institute major reforms of the revenue structure of BART and MUNI. In the short-term, yes—we have to vote for the tax. There is no immediate alternative that saves us from a generational transit disaster. In the medium-term, BART and MUNI need more counter-cyclical sources of revenue. Fare dependence and unstructured transfers made both systems uniquely vulnerable to ridership downturns; more structured funding mechanisms that we’ve mentioned above would alleviate this issue. And in the longer-term, BART and MUNI need to concentrate capital expenditure on automation, so it can break its cycle of perpetually rising costs.

Appendix

A: What About Caltrain?

You may have noticed the conspicuous absence of Caltrain from this article. This isn’t an attempt as obfuscation as much as one of fait accompli: Caltrain already passed its own sales tax back in 2020. Measure RR (confusingly, different from the BART “Measure RR” mentioned earlier) was a 0.125% sales tax in San Francisco, San Mateo, and Santa Clara, that established a dedicated revenue stream for Caltrain. This tax accounts for 60% of Caltrain’s revenue.

Before the pandemic, Caltrain was even more fare-dependent than BART, and it faced an even steeper ridership collapse. The suburban employers it served were the first to announce long-term work-from-home, and its lower density surrounds made driving a more substitutable mode for riders. The low-density also meant its proportion of non-commute usage was extremely low, to the point that pre-electrification in 2024, it only ran one train per hour on weekends.

But Measure RR was its stopgap non-fare funding source, and it was passed before the full crisis hit. That measure averted the scale of disaster that now threatens BART and MUNI.

B: The (East) Asian Question

East Asian transit systems, like Tokyo and Hong Kong, which are brought up as frequent comparisons to their inferior North American brethren, operate on an entirely different model. Tokyo, and especially Hong Kong, have transit companies as full landowners of their stations. These companies can build housing on their own land, manufacturing their own demand, and creating a constant growth flywheel: transit expansion creates housing, encouraging density, leading to job growth in the city, leading to further demand and more transit. This requires weak zoning laws, by-right permitting, and low construction costs, none of which exist in America. Elon Musk is going to launch data centers into space because permitting solar farms in the Mojave is an intractable delay; I doubt building by-right transit-oriented apartments in city centers will somehow be an easier permitting problem to solve.