TAHANAN, AT 833 BRYANT STREET in the Soma neighborhood of San Francisco, is 145 studio units of permanent supportive housing for the chronically homeless. Completed in 2021, it’s a cheerful, efficient building that bears the hopes and scars of the population it serves. The curated murals and architectural flourishes are pockmarked by extensive water damage inflicted when a resident on an upper floor reportedly slept with the faucets running. Social workers stride purposefully through the halls, and well-loved dogs are being walked everywhere you turn.
But what makes Tahanan notable isn’t its aesthetic. It’s the way it was built. Tahanan went up in three years, for less than $400,000 per unit.1 Affordable housing projects in the Bay Area routinely take twice as long and can cost almost twice as much. “Development timelines for affordable projects in San Francisco have typically stretched to 6 years or longer and development costs have reached $600,000 to $700,000 per unit,” reported the Terner Center for Housing Innovation at the University of California, Berkeley.2 San Francisco cannot dent its housing crisis at the speed and cost at which it is building affordable units now. But if the pace and price of Tahanan were the norm, the outlook would brighten.
So how did Tahanan do it? The answer, for liberals, is depressing: It used private money to avoid the pile of rules and regulations that taking government money triggers. But it could only do that because it had the support of city and state officials who streamlined zoning and cut deals to make it possible. Tahanan reveals a confusion in the way we talk of the government. The government is a plural posing as a singular. Different factions and officials and regulations and processes push in different directions. It is often the case that no one is more frustrated by how the government works than the people who work in it or who are charged with running it.
Tahanan was built on the former site of a parking lot and temporary bail bond office. Sounds easy enough to build on. But it wasn’t zoned for affordable housing. The project could get off the ground only because of legislation passed by State Senator Scott Wiener in 2017 that fast-tracked certain kinds of affordable housing projects in California past the local approval process.3 “This project didn’t have to go before the planning department for discretionary review or the Board of Supervisors,” Rebecca Foster, the chief executive of the Housing Accelerator Fund, which led the development of Tahanan, said. “We got our entitlements in four months, which is unheard-of.”
But that merely means you can begin the process of building. When you’re building affordable housing, you’re typically using public money. When you’re using public money, you have to abide by public requirements. Take the Local Business Enterprise and Non-Discrimination in Contracting Ordinance, also known as 14B.4 These requirements began in 1984 as a preference for minority- and female-owned contractors. But in 1996, California passed Proposition 209, which held that “the state shall not discriminate against, or grant preferential treatment to, any individual or group on the basis of race, sex, color, ethnicity, or national origin in the operation of public employment, public education, or public contracting.”5
Instead of scrapping the contracting requirements, San Francisco rewrote them to favor small businesses. “The public has an interest in fostering a strong and vibrant network of small and very small micro businesses in San Francisco,” the ordinance says. To qualify as one of the favored “Micro-Local Business Enterprises” under 14B, a contractor must have less than $12 million in average annual gross revenue.6 This cap creates a few problems. One is that it means public housing efforts in San Francisco are, by definition, discouraged from working with large contractors that have grown in size and revenue precisely because they are good at delivering projects on time and under budget. Another is that San Francisco has a tight labor market and an even tighter construction market. There aren’t a lot of capable small contractors sitting around with nothing to do.
In practice, Foster said, a few small contractors end up attached to a large number of affordable housing jobs, causing delays and cost overruns. Then, of course, there’s the cost of compliance—of proving to the city you’re following the 14B rules. Foster’s team estimates that requirements like 14B could add six to nine months and millions of dollars to building an affordable housing project the size of Tahanan.
It’s not just 14B. There are local hiring requirements. The Arts Commission does a separate review of your design. You need an additional review from the Mayor’s Office on Disability. Who could oppose that? But these projects already have to comply with the Americans with Disabilities Act, and the additional review takes time and comes at a cost. “They come in when you’re done,” Foster said. “And they’ll say, ‘That threshold is two centimeters off, and it is in all of your doors.’ And so that delays people moving in for another couple of months. And it might mean that you miss a financing deadline and have an adjuster on your tax credit fees that are another $2 million. So it just has a big ripple impact.”7
Tahanan is the first affordable housing project in San Francisco built using modular housing. All the units above the ground floor were fabricated at a factory in Vallejo, California. “That definitely helped with meeting the time- and cost-saving goals,” Foster said. But some local unions were furious, even though the factory in Vallejo is unionized. Here, then, is another place where progressive goals conflict. Local union jobs are a good thing. Modular housing can make construction cheaper and faster in a state facing a severe housing shortage. Which do you choose?
What made Tahanan possible was a $65 million grant from Charles and Helen Schwab.8 The grant’s conditions were that the housing had to be built in under three years and for under $400,000 a unit. By using private financing, the project avoided the standards and rules that public money carries. That isn’t to say the political system in San Francisco was against the project. The Board of Supervisors approved a crucial lease to keep the development operating into the future. But private money was the secret sauce.
It is damning that you can build affordable housing so much more cheaply and swiftly by forgoing public funds. Shouldn’t things happen faster when they are backed by the might and money of the government?
We are used to understanding the battle lines of American politics as cleaving liberals who believe in a strong, active government from conservatives who doubt it. The truth is far more complicated. Liberals speak as if they believe in government and then pass policy after policy hamstringing what it can actually do. Conservatives talk as if they want a small state but support a national security and surveillance apparatus of terrifying scope and power. Both sides are attached to a rhetoric of government that is routinely betrayed by their actions. The big government–small government divide is often more a matter of sentiment than substance.
Neither side focuses on what scholars call “state capacity”: the ability of the state to achieve its goals. Sometimes that requires more government. Sometimes it requires less government. But it always requires a focus on what the state is trying to achieve and what is in its way. In the absence of that focus, absurdity reigns.
Across Europe, government-administered health-care systems negotiate down the prices of drugs and treatments. In America, our fear of socialized medicine has led to a hodgepodge of private and public insurers who do not coordinate and do not effectively negotiate. The weight-loss drug Ozempic, for instance, costs about ten times as much in America as it does in Britain or France.9 Those countries have national health-care systems that restrict what pharmaceutical companies can charge, and we do not. As a result, taxpayers in Europe spend less on health care, as a percentage of GDP, than taxpayers in America. And then Americans have enormous private bills atop our public spending. Keeping the American health-care state weak has made the American government larger and left Americans poorer.
But liberals lose sight of their goals, too. In response to the Tahanan story, Bob Kuttner, the cofounder of the stalwart liberal publication The American Prospect, tried to jam the problem into a container that the left is more comfortable with. “We have a very modest social-housing sector in the US and limited funds for housing subsidies. We are largely at the mercy of developers. We could eliminate zoning restrictions and make it easier to build multifamily housing, and that would solve only a small portion of the affordable-housing shortage.” There is a comfort here with solutions that put more faith in government (public, or social, housing) and a discomfort with solutions that seem to align with markets (being at the “mercy” of developers, eliminating zoning restrictions).
The reality of housing development doesn’t track along such neat ideological lines. Kuttner says that eliminating zoning restrictions and making it easier to build multifamily housing would make only a modest difference in our problems. He does not provide any evidence for this claim, but there is evidence against it. Houston has no zoning rules at all, though it does have some land use regulations.10 As a result, it is dramatically easier to build in Houston than to build in Los Angeles or San Francisco or Seattle or Boston.
In 2023, the San Francisco metro area issued about 7,500 new housing permits. The Boston metro area issued 10,500. New York City, Newark, and Jersey City—together—issued slightly fewer than 40,000. The Houston metro area issued almost 70,000.11 This divergence is decades old, and its consequences are clear. Houston has the lowest homelessness rate of any major US city. Officials estimate that it costs $17,000 to $19,000 to house a homeless resident of Houston, with about $12,000 of that going to housing and the rest to wraparound services.12 In San Francisco, the cost is between $40,000 and $47,000 annually, with about $35,000 going to housing costs alone. This tracks the broader difference between the two cities: in Houston, the median home costs a bit over $300,000 rather than a bit over $1.7 million in San Francisco.13 Houston is not free of affordability problems. But it is not facing the crises of homelessness and housing affordability seen in the superstar cities of many blue states.
Liberals lament that private developers want to build profitable developments when what is needed most is affordable housing. But even aside from how much housing is built, one way to make housing more affordable is to make it cheaper to build. The problem is many liberal jurisdictions have layered on rules and regulations that make housing pricier even when it is constructed—and that, of course, makes it less affordable. In San Francisco, a 2023 state report found that it took 523 days, on average, to get clearance to construct new housing, and another 605 days to get building permits—and that’s for the projects that aren’t killed by community opposition during the planning process.14 A project needs to be quite profitable to make it through that gauntlet—and it needs to be acceptable to its wealthy neighbors—and that pushes developers toward luxury condos.
But the grim absurdity of liberal housing policy comes clearest when you focus on the kind of housing liberals claim to support: affordable housing built by nonprofit developers with the backing of both voters and local government. In 2016, the people of Los Angeles overwhelmingly passed Proposition HHH, a ballot measure that raised $1.2 billion through a higher property tax to create 10,000 new apartments for the homeless. “The voters of Los Angeles have radically reshaped our future,” Mayor Eric Garcetti said, “giving us a mandate to end street homelessness over the next decade.”15
By March 2024, the city had built 4,344 units under HHH.16 A 2022 audit found the units cost, on average, around $600,00017—almost twice the cost of the median sale price for a home in Houston. There have been many problems with Prop HHH, but the real problem predates it: the way that taking advantage of public money layers on requirements, delays, and additional goals, slowing down construction and raising costs.
HHH is designed to provide some, but not all, of the money for developments. Defenders of HHH are quick to point out that the average cost per unit includes around only $134,000 of HHH funds.18 The program is designed to seed projects that can find other financing, too. That sounds good: by leveraging outside money, the taxpayer’s dollar can go further. In reality, it means affordable housing projects need to line up four or five or six different funders, cobbling together tax credits and philanthropic donations and state and local incentives.
“Everyone wants to be able to say we spent only $50,000 on this apartment—that means I have to go through the process four or five times,” Yasmin Tong, the founder of CTY Housing, a consultancy on affordable housing projects, says. “I’ve seen projects with as many as ten funding sources. It takes time to do that.”19
The different financing sources come with different demands, all of which make the project more complex. “The developer has to hopscotch from one funding source to another to another,” Tong said. “So you start by saying we’ll serve low-income families at this development. But the funding falls through. So you put in veterans’ units. Or try to house domestic violence survivors in here. There’s this constant restructuring of the project as the funding sources come and go.”
Ron Galperin was the Los Angeles city controller from 2013 to 2022. His office was responsible for auditing HHH. His office tracked how much each unit cost and where the money came from and whether the program was achieving its aims. You might expect him to praise HHH’s effort to match each dollar with five dollars from other funders. In fact, he’s furious at the way the money was structured.
“If you look at the inflated cost that comes along with all of the regulations and rules and restrictions and limitations,” Galperin said, “then basically all of this money is going to feed the beast of covering the cost of the regulations. Yes, they get $134,000 on average from the city, but the hoops that have to be jumped through to get it very well may exceed the $134,000. We’ve created an absolutely insane system.”20
Then there are the higher standards that public money requires developers to meet. “We’re required to pay prevailing wage, so there’s at least a 20 or 30 percent premium on the labor costs,” Tong says. “We have sustainability requirements we need to maintain. I’ve had projects where the planning department required a higher-quality air ventilation system because we were a certain distance from the freeway. All the affordable housing development is subject to green building requirements. The standards in California are higher than anywhere else in the country. And you’re not just required to build to the standard, you also need to hire a consultant to confirm you’ve built to the standard. That adds costs.”21
Every one of these is a worthy goal. But so too is building a lot of affordable housing quickly and cheaply. Los Angeles is failing, and failing badly, at doing that. Given that failure, does it make sense to be asking for special air filtration systems for developments near freeways when the alternative, for many of the would-be residents, is a tent beneath the freeway? To pose the question sounds callous. But to refuse to pose the question, given the need for more housing, is cruel.
These additions do not come only from planning boards trying to upgrade the quality of the housing. They also come from neighbors who would prefer it never got built at all. In Venice, home of the legendary boardwalk, the Venice Dell Community project is trying to turn a parking lot owned by the city into a 140-unit building for homeless residents, low-income artists, and families, all of it designed by a star architect.22 The development is being fought and even sued by a collection of local homeowners who complain that “Venice desperately needs this parcel to address our chronic parking shortage,” that the new housing would be “an eyesore completely divorced from sound architectural principles,” and that it is being developed “with no environmental review in a designated tsunami zone and FEMA Special Flood Hazard Zone.”23 (When do Angelenos want affordable housing? Now! Where do they want it? Not here!)
Surviving local opposition often means agreeing to a range of demands that send costs ballooning. To try to neutralize local attacks, developers hire pricey architects, redo plans repeatedly, make all kinds of aesthetic and architectural concessions or additions, hire extra lawyers and auditors, and on and on. Even if a project does survive all this, it does so at a higher per-unit cost, which then, of course, becomes one more data point that gets wielded in opposition to the next project.
Perhaps, as Kuttner suggests, the problem is simply that we don’t have enough public (or, as it’s been rebranded, “social”) housing. In Singapore, almost 80 percent of the population lives in public housing. These projects have a bad reputation in the United States, but beautiful developments have opened in places like Montgomery County, Maryland, and larger cities like Atlanta are experimenting with using public projects to expand their housing stock. But social housing will rise or fail for the same reasons that all building projects rise or fail. It doesn’t matter whether the worker hammering in nails is a public employee or a private contractor. The government still needs to build those homes affordably and quickly. And that’s not possible under the rules and strictures that liberals have designed within the governments they run.
Heidi Marston led the Los Angeles Homeless Services Authority from late 2019 until April 2022, when she resigned in frustration. “We had thirty-eight unique funding sources coming in when I was there,” Marston says, “and each of those had annual or biannual audits of not just us, but the nonprofits we were funding.” Those audits were meant to show that the money was being spent exactly as intended. But that was part of the problem.
“Federal funding is probably more restrictive than any other,” Marston continued. “Every year we get money from the Department of Housing and Urban Development. The city often gives their share to us, but on top of the auditing and tracking that the federal money comes with, they add on their own conditions, like we can’t use it for staffing. Just all this stuff that gets added on in the process.”24
It is easy enough to imagine how these conditions emerge. The city wants to show that it is using its money to build houses rather than expand its head count. HUD has no end of priorities and is trying to satisfy the desires and demands of the members of Congress who control its funding. Tax credits are added to the code to address real and wrenching problems, like the rise in homelessness among veterans. Grant makers want to show donors that their money is being used well, and the only way to prove that is through audits. Everyone, everywhere, is afraid of being implicated in fraud or waste or having their funding cut or seeing the public turn on them.
Each individual decision is rational. The collective consequences are maddening. We hire skilled, dedicated people to do the public’s work and then make it impossible for them to do that work well. We ask people to work on society’s hardest problems—often making much less than they could make in the private sector—and then rob them of the discretion and agility they need to solve them. And then we wonder why so many of them leave.
“There’s tons of money that goes into homelessness, particularly in Los Angeles,” Marston says. “My budget was almost a billion dollars. But the money comes with such confined requirements that it’s almost impossible to spend. If you give me a billion dollars and the ability to spend it, it would be a different story.”25
It is hard to hear Marston’s story without being reminded of Nicholas Bagley’s argument that liberalism has become obsessed with procedure rather than with outcomes, that it seeks legitimacy through rule following rather than through the enactment of the public’s will. Homelessness in Los Angeles is a catastrophe. The public is furious at the sluggish, ineffective response. And the lead agency on homelessness is spending its time filling out audit forms and making sure each dollar is spent in strict accordance with the specific demands of funders.
In his 2022 article “A Time for Triage,” Michael Gerrard, the founder of Columbia Law School’s Sabin Center for Climate Change Law, considered why it has proven so hard for liberals to build the kind of climate infrastructure they believe is needed. “Rather than climate denial, the environmental community has tradeoff denial. We don’t recognize that it’s too late to preserve everything we consider precious, and to linger in making decisions. Society has run out of time to save everything we want to save, and to mull things over for years.”26
One problem liberals are facing at every level where they govern is that they often add too many goals to a single project. A government that tries to accomplish too much all at once often ends up accomplishing nothing at all. (Conservatives are not immune from piling on procedure and stricture, but they often do so in a purposeful attempt to make government work poorly, and so failure and inefficiency become a perverse form of success.)
Call this “everything-bagel liberalism.” The everything bagel is, of course, the best bagel. But that is because it adds just enough to the bagel and no more. It does not, actually, pile everything atop the bagel. In the Oscar-winning movie Everything Everywhere All at Once, there is an attempt to create a true everything bagel, and it becomes a black hole from which nothing can escape. The same is true for public projects. When the government adds the right number of goals, standards, and rules, much can be accomplished. When it adds too many, the project can collapse under its own weight, as has happened to high-speed rail in California.
In 2022, President Biden signed the CHIPS and Science Act into law. The Biden administration believed semiconductors would be to the twenty-first century what oil was to the twentieth century and that America must be a leader again in manufacturing them. “This is first and foremost and primarily a national security initiative,” Gina Raimondo, the secretary of commerce, said. “We have national security goals we must achieve. Period. Full stop. No compromise.”27
The semiconductor industry was invented in America—the “silicon” in Silicon Valley refers to the material that semiconductors are made from—but we long ago lost our dominant position in making what we invented. A report by the Semiconductor Industry Association says that the US share of global semiconductor-manufacturing capacity dropped from 37 percent in 1990 to 12 percent in 2020.28 Part of the reason is cost. The association estimates that building and operating a semiconductor-manufacturing facility in the United States costs about 30 percent more over ten years than it does in Taiwan, South Korea, or Singapore.29
In 2023, the Biden administration released its Notice of Funding Opportunity for the $39 billion it intended to hand out to semiconductor manufacturers to locate new fabs in the United States. Reading the NOFO was a strange experience. Here was the US government trying to recapture an industry it had lost in part because it had become cost-prohibitive to manufacture semiconductors domestically. But the NOFO did not seem laser-focused on the cost problem. To be honest, it did not seem laser-focused on any problem.
Page 12 encouraged a pre-application that includes an environmental questionnaire “to assess the likely level of review under the National Environmental Policy Act.” Page 20 mandated that applicants prepare “an equity strategy, in concert with their partners, to create equitable work force pathways for economically disadvantaged individuals in their region,” which should include “building new pipelines for workers, including specific efforts to attract economically disadvantaged individuals and promote diversity, equity, inclusion, and accessibility.” Pages 21 and 22 asked for a plan “to include women and other economically disadvantaged individuals in the construction industry” and encouraged the use of project labor agreements and sets out requirements for “access to child care for facility and construction workers.”
Pages 24, 25, and 26 asked applicants to detail how they would include minority-, veteran- and female-owned businesses, as well as small businesses, in their supply chain, and offered seven bullet points detailing how this might be done, including dividing supply chain requirements “into smaller tasks or quantities to expand access” and “establishing delivery schedules for subcontractors that encourage participation by small, minority-owned, veteran-owned, and women-owned businesses.” Then there are requirements for “a climate and environmental responsibility plan,” as well as community investments in areas like transit, affordable housing, and schools.30
Many of these are good goals. But are they good goals to include in this project? There is no discussion in the NOFO of trade-offs. Nor was there any admission by the administration that anything they were asking for even represented a trade-off.
“Every one of the requirements—or they’re not really requirements—nudges—are for criteria or factors we think relate directly to the effectiveness of the project,” Raimondo said. “You want to build a new fab that will require between 7,000 and 9,000 workers. The unemployment rate in the building trades is basically zero. If you don’t find a way to attract women to become builders and pipe fitters and welders, you will not be successful. So you have to be thinking about child care.”31
But do Taiwanese semiconductor firms really know how to expand the role of women in the construction industry? How good will they be, really, at diversifying supply chains? These are all worthwhile goals. But there is some margin at which trying to do more means ultimately achieving less.
It is impossible to read these bills and guidelines and not notice that the additions are rarely matched by deletions. Process is enthusiastically added but seldom lifted. You can imagine a version of the CHIPS bill that lifted immigration rules to make it easier for skilled semiconductor workers to come to the United States. That would have been the most direct way to address the shortage of skilled workers hindering the construction and operation of the fabs. You could have imagined rules exempting the semiconductor fabs from NEPA or giving them some kind of fast-track process. (In late 2024, the Biden administration signed a bill from Senators Mark Kelly and Ted Cruz to do exactly that, after warning that environmental review could add “years” to the construction timeline.)
To be clear, there is nothing unusual in the way the Biden administration approached the CHIPS and Science Act. The federal government often tries to make the subsidies it offers serve an array of goals and constituencies. California’s high-speed rail was shaped by this dynamic, too. Many Californians were confused that construction had begun in the Central Valley, which was far less populated than the corridors near Los Angeles or San Francisco.32 Why start there?
When California applied for federal money under the terms of the American Recovery and Reinvestment Act, the Obama administration gave preference to bids that would improve air quality in poor communities. And so the $3 billion the federal government offered was not really to build high-speed rail. It was to begin building high-speed rail in ways that addressed air pollution in specific places. The Central Valley is poorer and more polluted than coastal California, so federal funding went there, and so did the initial construction. California is building high-speed rail in a place that makes it less likely that it will generate the ridership, political support, and financial backing to ever finish. The irony is that it’s not just bad for the high-speed rail project. It’s also bad for air pollution across the state.
Since 1960, federal government spending has risen more than fivefold—and yes, that’s accounting for inflation.33 But the size of the federal civilian workforce has barely budged. It was slightly fewer than 2 million people in 1960 and it’s slightly over 2 million people today. In countries like China and Singapore, civil service is held in high esteem, and the brightest graduates compete in nationwide tests to win government jobs. In the United States, the word “bureaucrat” is tossed around as an epithet. Republicans have spent decades demonizing government, and they have largely won the argument. The dominant belief is that anything that can be outsourced or privatized should be. Government is bloated. The private sector is efficient.
Democrats may not believe what Republicans believe about government, but they often act as if they do. In 2008, when California began building its high-speed rail system in earnest, the state’s High-Speed Rail Authority had just ten workers. One of them was responsible for designing graphics for social media. The job was turned over to a vast assemblage of consultancies. It was one of these consultants—WSP—that estimated the system would cost only $33 billion and take only twelve years to build. But WSP was joined by Project Finance Advisory, Cambridge Systematics, Arup, TYLin, HNTB, PGH Wong Engineering, Harris & Associates, Arcadis, STV, Sener, and Parsons Corporation. The outsourcing “proved to be a foundational error in the project’s execution—a miscalculation that has resulted in the California High-Speed Rail Authority being overly reliant on a network of high-cost consultants who have consistently underestimated the difficulty of the task,” reported Ralph Vartabedian in the Los Angeles Times.34
California is one of the richest polities in the world. It was building one of the most ambitious rail projects in the world. But it did not hire the best rail designers and engineers to provide in-house expertise and manage the project. California was financing and overseeing a program it did not have the capacity to plan, manage, or even truly understand. “There was an ideology at the Authority some time ago that was like, ‘Let’s keep this small and in-house and we’ll rely on consultants to build this,’ ” Brian Kelly, the High-Speed Rail Authority’s CEO, said. “My philosophy when I got here was the state is the owner of this project and so we need to build state capacity. When I started, the authority was seventy percent consultants, thirty percent state. Now it’s fifty-five percent state and forty-five percent consultant.”35
In the Bay Area, a different story played out. In 2012, Bay Area Rapid Transit (BART) signed a contract with Alstom, a French rail car manufacturer, to deliver 775 cars for $2.58 billion.36 By 2023, though, something unusual had happened: the cars were coming in faster, and cheaper, than expected. The cost estimate was slashed by almost $400 million.37 One major source of savings, reported trains.com, was “BART’s decision to have its own staff do more of the engineering work in house. The project team has included engineers who have successfully completed new rail car projects at other agencies.”38 Nor is this an isolated anecdote. Zachary Liscow’s research found that increasing employment in state departments of transportation by 1 employee per 1,000 residents reduced the cost-per-mile of highway construction by 26 percent.39 Government cannot do everything itself. But it needs enough know-how to oversee the projects it is doing.
Jen Pahlka is the founder of Code for America, a civic tech group that tried to build a bridge between the technology industry and the government in a bid to upgrade government services. The work was hard, fruitful, important, frustrating. She went on to advise on digitizing government in the Obama White House. She is something of a godmother to a generation of idealistic technologists who tried, or are trying, to make government work the way it should work. Her memoir of this work, Recoding America, is a compendium of their stories. It is painful to read.
In January 2020, Pahlka had stepped away from her role at Code for America. She needed some time away from the problems of digitizing government. But then came the pandemic, and the lockdowns, and the millions of people suddenly out of work. Those people were all, unexpectedly, now reliant on unemployment insurance, which is managed by the states. And those systems were not prepared for anything like this level of demand. California’s system, administered by the state’s Employment Development Department, fell into particular chaos, with millions of people seeing their benefits wrongly delayed or denied.40 Pahlka was asked to co-lead a task force that would rescue it.
Technologically, there was nothing particularly novel in the challenge. Unemployment insurance is fairly simple. People apply. They are accepted or rejected. Then checks are sent out. By the standards of the technology sector, this is a solved problem. “Privately, some California officials told me they thought the EDD staff was just incompetent at technology and our team would find the problems easy to fix,” Pahlka writes. But that wasn’t how the task force saw it. “Privately, we wondered if we could help at all.”41
Pahlka has come to think of government technology—and the regulations that control it—as layers of sediment. As new problems emerge, new layers are added. But the older ones are rarely removed. “Each successive layer is constrained by the limitations of the earlier technologies,” she writes. “The system is not so much updated as it is tacked on to.”42 The challenge of updating government technology is the challenge of updating, harmonizing, or terminating the functions of these old systems. And all of it must be done while following procurement and contracting rules that no private technology company would ever impose on itself.
At the EDD, the core technological layer was called the single client database, which runs on an IBM mainframe from the ’80s.43 Parts of it are written in a programming language called COBOL, which dates back to 1959. COBOL is almost never used today, and it is hard to find engineers who know how to program in it. Making matters worse, parts of the single client database were designed to run on those old monochrome displays that showed green text on a black background. Because nobody makes those displays any longer, the staff used virtual emulators to access the system—they would run software on new computers that could mimic the constraints of old computers.
Then came more layers. In 2002, the EDD contracted with Deloitte to bring their work online. Deloitte built one system to access the IBM mainframe through a web browser. It built another system to corral and manage applications flagged for manual identity verification. It built a third system that acts as the public-facing website for people to apply for benefits. All these systems had their own subsystems. And within those subsystems, applications could pool and get trapped in places no one was really looking. Pahlka and her team were told the number of backlogged applications was around 230,000. It took them seven weeks to organize the databases such that they could be precisely counted. The true number was 1.2 million.
The EDD doesn’t build or manage its own technology. Nor is that technology built or managed by a centralized team of software engineers in the state government. It is done by external firms chosen and managed through a labyrinthine procurement process. At the time of the meltdown, the EDD had been working on a modernization contract for ten years that it was theoretically just weeks away from awarding. Read that again: They had not been working on modernizing their technology stack for ten years. They had been working for ten years on the massive contract they would award to outside firms to modernize and manage their technology stack. That contract was expected to take eleven years to execute.
The sedimentary chaos at the EDD was not at all unusual. California spent ten years and $500 million trying to bring its courts onto a common document management system before abandoning the effort.44 The State Department’s Bureau of Consular Affairs has been trying to modernize and consolidate its visa and passport systems since 2009.45 The IRS began trying to replace one of its core systems—the Individual Master File—in 2000. The work is now projected to be completed in 2030.46
“The public servants responsible for the interminably drawn-out modernization efforts are neither lazy, stupid, nor malicious,” Pahlka writes. “I’ve met hundreds of them, and they are overwhelmingly dedicated, conscientious, and often quite creative. IRS employees managed to send monthly child tax-credit payments to nearly forty million families and to mail out over $800 billion in stimulus checks during the pandemic, all while relying on systems that were never designed to change so quickly or handle such enormous volume.” The problem is that the systems they are updating have become “complex beyond our ability to imagine,” as has “the complexity of all the rules these public servants need to follow to do that updating.”47
The worst of the EDD backlog was in the system that managed manual identity verification. But working in that system required years of experience, accreditation, and testing. When the EDD crisis had begun, elected officials demanded the EDD hire more people. So the EDD signed another contract with Deloitte to bring on another five thousand workers. The governor touted the new hires. But it would have taken years to train those workers to face down the backlog the EDD was facing. And their questions and confusion were taking up the time of the workers who could work on the backlog. Pahlka’s team calculated that it was now taking two to five times as long to clear those files as it had before the pandemic.
Letting go of thousands of new hires is cheaper and easier than training them. But that wasn’t how the agency’s leadership saw it. “Hiring as fast as they possibly could had been the one consistent directive coming from everyone above them: the governor’s office, the legislature, the federal Department of Labor, and every oversight body with jurisdiction over the EDD’s operations,” Pahlka was told.48 Telling all those overseers they were wrong was not in anyone’s interest. And no one believed they would listen anyway. Firing workers during a crisis of EDD performance would look terrible.
There was another option. The system was choking over manual verification. Manual verification was typically triggered when the information an applicant filled out on their form didn’t precisely match some other piece of information the EDD had about them. Perhaps you write “Jonathan” on legal forms but your employer pays you under “John.” Perhaps you mistyped a digit in your Social Security number. It makes sense why this would lead to a manual check. But there was no real relationship between these tiny errors and fraud. Out of the 183,167 claims flagged in the previous quarter, only 804 were ultimately judged invalid.49
If anything, there was more fraud in the perfect applications. “Our world is awash in databases of stolen identities from breaches at credit monitoring services, retailers, and employers, and these stolen identities are freely traded on the dark web. Fraudulent applications using these sources will not get flagged: the data entered on the application will exactly match the sources the EDD checks against, because it is usually a copy of precisely that data,” Pahlka writes.50
The EDD was implementing a new system of identity verification that would be quicker and more effective, but it needed to do something about the backlog that was building daily. The obvious answer was to loosen the rules that would lead to manual verification. But even though the process wasn’t working, it was still the process. To follow it was safe. To evade it was risky. Fraud was really happening, and when its full extent was known, there was going to be a furor, and it would fall particularly heavily on anyone who loosened the anti-fraud rules, even if the rules they loosened were failing to catch fraud and causing the huge backlogs that were crashing the system.
What the EDD eventually did was simply stop taking applications altogether. For weeks, they shut down the portal for new applications. The EDD reassigned the bulk of its staff to clearing the backlog and setting up the new identity verification program.
Amid all this, Pahlka recalls, a member of the California Assembly introduced legislation requiring the EDD to make its applications and communications available in over a dozen languages. Most of those languages were already required by a 1973 state law. They were also required by multiple federal laws and rules. The EDD wasn’t in compliance with all these older rules. It wasn’t even serving English speakers effectively. It was not able to do what it was already required to do. Now it was being instructed to do more.
What was needed was subtraction. What Pahlka and her team found, again and again, was that the rules and regulations that governed California’s unemployment insurance system and that had been written into its code had just kept growing. That made the code more complex and harder to update. It made new hires harder to find and harder to train. It made backlogs harder to clear. “Lawmakers often have good intentions, but they continually add policy layers with too little understanding of (and, sometimes, regard for) how what they add will interact with the layers that are already cluttering the delivery environment,” she concluded.51 For government to do more—or even for it to just do what it is already doing—sometimes it first needs permission to do much less.
On June 11, 2023, a tanker truck carrying 8,500 gallons of gasoline flipped over. The truck ignited underneath the I-95 bridge in Philadelphia, killing the driver and melting the steel beams undergirding it. The I-95 bridge, which carries 160,000 cars daily, collapsed. This wasn’t just a crisis for a roadway. It was a crisis for a region. I-95 is one of the main transportation arteries on the East Coast. It’s a crucial connector between New York and Washington. Officials, including Pennsylvania governor Josh Shapiro, warned that rebuilding it would take months.
And it would have taken months, or longer, under Pennsylvania’s normal rules. “We would hire a consultant to design it,” Mike Carroll, the Pennsylvania secretary of transportation, says. “We’d need final design approved by the Federal Highway Administration. Then there’d be bidding from interested contractors. Then we’d process the bids. Then we’d issue a contract. That’d take about twelve to twenty-four months.”52
But Shapiro signed a declaration of emergency that exempted the rebuilding process from the rules and requirements that slow so many public projects down.53 Speed was the priority here. There would be no environmental impact statement. There would be no lengthy bidding process. The procurement rules were shunted aside. When Carroll arrived at the disaster, C. Abbonizio Contractors, a firm the state had worked with before, was already at the bridge on another job. They were chosen to oversee the demolition. Rob Buckley, of the highway contractor Buckley & Company, was also nearby, working on another project. His firm was pulled in, too. “The emergency declaration gave us the ability to engage contractors without bidding,” Carroll said. “Work commenced the moment the fire department released the scene—that same day.”
All the labor Pennsylvania used was union labor. And they pushed hard: work went on twenty-four hours a day, seven days a week.54 A twenty-four-hour live cam trained on the site allowed the public to follow along. Shapiro took to giving updates on Twitter and TikTok. He turned the I-95 rebuild into a crucible for his governorship and an object lesson in something few still believed: That government could build big things fast. That it could do so using union labor. That it could move at the speed of an emergency rather than according to its own rules.
“The common denominator with all these decisions was let’s get this thing done as fast as possible,” Carroll said. He recalled a moment he came across a bunch of Abbonizio workers using a screwdriver to disassemble a highway sign. He asked what they were doing, and they told him they were saving the sign for the Department of Transportation in case they wanted to reuse it. “I said turn the machine on and knock the goddam thing over,” he said with a chuckle. On another night Carroll saw rain forecast over the next few days. He told the team to pave anyway, in contravention of the Department of Transportation’s rules, because the rain was light that night and could grow heavier soon.55 If they waited, they might be waiting for days. “The emergency declaration was a game changer,” Carroll says. “I took calculated risks that I’d have not taken in a normal project. It could’ve gone badly, but it didn’t.”
It’s worth taking seriously what Carroll says there. These were risks. There are reasons these rules are in place. No-bid contracts can enable corruption as well as speed. There are reasons not to put down asphalt when it’s raining. But in turning these questions from choices into rules, we have taken discretion and judgment away from people like Carroll. We prefer that projects go badly by the book. We minimize some risks but make delay and high costs routine.
The emergency declaration allowed Shapiro to make choices. He chose to use union labor but to gore a lot of other interests and processes. I-95 reopened in just twelve days—not the “months” initially forecasted. Shapiro did “one heck of a job,” President Biden said.56 His popularity swelled, and he began to be mentioned as a possible future presidential candidate. Turns out people like it when their government gets things done.
The Washington Post asked Shapiro to write an op-ed reflecting on lessons he’d learned. The first lesson, he said, was “empower strong leadership.” The key to the rebuild was that the people in charge of the rebuild could act. “Managers of every component of the project were empowered to be decisive, take ownership and make a call when necessary—not defer and delay to the often-circular bureaucracy,” Shapiro wrote.57 The process Shapiro used would typically be illegal. Yet national Democrats and Pennsylvania voters alike loved it. What does that say about the typical process?
In his paper “State Capacity: What Is It, How We Lost It, and How to Get It Back,” Brink Lindsey puts it well:
What is needed most is a change in ideas: namely, a reversal of those intellectual trends of the past 50 years or so that have brought us to the current pass. On the right, this means abandoning the knee-jerk anti-statism of recent decades, embracing the legitimacy of a large, complex welfare and regulatory state, and recognizing the vital role played by the nation’s public servants (not just the police and military). On the left, it means reconsidering the decentralized, legalistic model of governance that has guided progressive-led state expansion since the 1960s, reducing the veto power that activist groups exercise in the courts, and shifting the focus of policy design from ensuring that power is subject to progressive checks to ensuring that power can actually be exercised effectively.58
Liberals have chosen to trust elected politicians and government workers less and trust regulatory and judicial processes more to ensure that government delivers. That may have made sense in a past era, but given the problems we face now, it is a mistake. Whether government is bigger or smaller is the wrong question. What it needs to be is better. It needs to justify itself not through the rules it follows but through the outcomes it delivers.