What the death of Direct File tells us about state capacity

We CAN have nice things - if we want them

Source: Till Lauer at the New York Times

“Taxes are what we pay for a civilized society” according to Chief Justice Oliver Wendell Holmes Jr.. But the billions of hours and hundreds of billions of dollars Americans spend on tax reporting are the price we pay for a dysfunctional government unwilling to invest in public interest capacity.

In addition to tariff- and war-induced inflation, Americans can also thank President Trump for allowing this time tax to grow. Trump killed Direct File, a free tax reporting option that IRS had built. Binyamin Appelbaum of the New York Times summed up how I and a lot of other people who care about government services feel:

He has destroyed things that are more important than Direct File, but this one sticks in my craw. It was a straightforward way to make life a little better for a lot of Americans. It was a step toward the kind of easy-to-use, efficient, high-tech government services that everyone claims to want. It worked. And now it’s gone.

What Appelbaum is pointing out that Direct File is not just about whether you save some money at tax time. It is also about whether American government can deliver modern public services in a digital age. One member of the Direct File team I spoke to noted their initial surprise to learn that the US was lagging so far behind other countries when it came to digital tax reporting:

I think that we should have greater ambition to actually do things on the Internet in our government. Just because we’re big and complex and multiple time zones doesn’t mean it’s not possible. So I kind of came with that attitude and was stunned, absolutely stunned to find out that the IRS didn’t offer a free tax filing thing.

Over the last few months I’ve been talking to more than 20 people who worked on the Direct File project, everyone from the top of the policy food chain — like former Treasury Secretary Janet Yellen and IRS Commissioner Danny Werfel — to the engineers and designers who worked directly on building the tool itself. This tax season, I’m sharing what I learned. For more detail, you can read a Better Government Lab working paper.

Direct File showed that government could build a high quality tax-reporting digital tool

The idea of reducing tax reporting burdens is not new. Presidents since Reagan made commitments to build something like Direct File before Biden delivered. The logic was clear: you should not have both pay taxes, and pay to complete your legal obligation to report your taxes.

You will never guess who also proposed an IRS app to help for Americans file their taxes.

When I asked those who helped build Direct File about its legacy, a consistent response is that they showed it could be done. Former IRS Commissioner Werfel said “it’s a proof point that the IRS itself can deploy user-friendly paradigm shifting technology in a short turnaround if the preconditions are set up for it to do so.”

The arguments about the technical feasibility about government provided tax-reporting are over. Direct File was small, but it was real and it worked and could have been expanded to tens of millions of taxpayers. Instead, it was killed under DOGE and Trump.

Users loved Direct File:

In 2024, 86% of users said it increased their trust in IRS, not the most beloved of organizations. The Direct File “net promoter score” (a measure of customer satisfaction and loyalty) is 84, well above that of TurboTax, and even Apple. In user surveys conducted by Code for America, 94% prefer Direct File and state equivalents to their previous filing method and 84% said they were “very satisfiedˮwith another 14% being “satisfied” with the product.

Design/Illustration by Shanima Parker / Scoop News Group

Why was Direct File killed?

Without the technical arguments, what are the reasons not to do Direct File? The Trump administration points out that it was not widely used. This is true, but the criticism is a little rich since the Trump administration made no effort to promote the new product it inherited.

The limited use also reflects a strategic choice by the Direct File team to take a “start small, fail small” approach. This was in response to the long shadow of Healthcare.gov, the Obama-era website that rolled out on day 1 without sufficient testing and immediately and very publicly collapsed. Direct File did the opposite — lots of testing, lots of discovery of where users found problems, lots of iterative fixes, improvements, and gradual additions of capability. This approach recognized the political reality that a single big failure would have killed the project. It also matches best practice with how to build a digital product.

It seems more than a little ironic, then, that Direct File was filled by people who said they were bringing tech innovations to government. The fate of Direct File under DOGE was initially unclear. As noted above, Musk talked about building a product that sounded a lot like Direct File, but he also pledged to get rid of the tech unit 18F when someone pointed out they were involved in Direct File

It could be that Musk and DOGE’s intolerance of technologists who did not come from their stable contributed to Direct File’s demise, despite the fact that the Direct File team was doing the sort of thing DOGE said it was going to do: use tech skills to build new products in-house, without relying on expensive contractors.

What we do know is that DOGE officials in IRS initially praised the Direct File team, and reassured them the project would continue — until right before it was cancelled. They changed their tune after meeting with representatives from the private tax preparation industry. Soon after, Direct File was killed.

Intuit, the maker of Turbotax, and other players in the private tax industry poured extraordinary resources into preventing any public option for tax filing. By one measure, Intuit had more lobbyists than the entire Direct File team had employees, and ramped up its spending as Direct File was launched.

The Trump administration says “Direct File had low overall participation and relatively high costs and burdens on the federal government, compared to other free filing options.” I literally wrote the book about administrative burden in the public sector and this has it completely backwards. It is a calculation that only makes sense if you do not value the time and money of citizens, but regard the government doing anything to help the public as a burden.

The average American spends about $290 to file. They also spend about 13 hours, often giving the government information it already has, which works out to another $486 worth of time assuming the average BLS hourly wage.

The Trump administration says that Direct File was too expensive. Compared to what? Direct File cost IRS $16 million in its final year of operation. Between 2021-2025 Intuit also got almost $500 million dollars in federal tax credits.

Public interest capacity in a democracy

And here I want to make some points about state capacity. I think Direct File tells us about capacity in a couple of ways. First, it reflects whether government can provide digital services that helps the public in a political economy where private sector actors can both make money from those services, and use some of that money to lobby the government to kill a public option.

The Trump administration is building state capacities in certain domains, such as immigration enforcement and surveillance. But if we think about public interest capacity as the ability of the state to generate classic public goods with widely distributed benefits, the Trump administration is largely failing. It is not an exaggeration to say that in order to facilitate a tax cut, your government added administrative burdens to the safety net so that both administrators and the public will spend more time and energy to get less and less from a dysfunctional system. They see value in breaking rather than fixing public services.

Public interest capacity matters to the credibility of democratic systems. The ability of government to competently help the public broadly reinforces liberal democratic systems. Historically, states with administrative capacity centered on merit and impartial public services are less likely to experience democratic breakdowns relative to governments driven by cronyism. The failure to deliver basic public services appears to fuel anti-government populism, even if the populists who benefit end up delivering worse services because they don’t take governance seriously.

The capacity to build digital services

The second capacity point is that Direct File showed us that government can change how it builds such services. The proof of concept that they offered was not just about tax reporting, but about different ways of operating. One team member said: “Direct File was also demonstrating a way of working to the rest of government. This was bringing the US government into the 21st century.”

For people who study or have worked in government there is an emerging conventional wisdom that current modes of operating simply do not work: they are dominated by contractors, cost too much and often don’t deliver. Making it worse, governments traditionally structure contracts in ways that did not allow iterative improvements, and do also not have internal capacity to manage complex tech products built by its contractors.

A member of the Direct File Team described the choice between building internal capacity and

the traditional IRS way: we create a bunch of requirements and we toss it over a fence to a vendor who then builds something and comes back to us in six months, and we see what the product is. We argued pretty hard that that’s not the approach we should be taking. And in fact, we should be building it ourselves in-house…to show that you can build good government technology that doesn’t have to be built by vendors, that it could be built internally within an agency…showing that you can actually transform how this agency works, helping to drive larger changes that would have impacts beyond just Direct File and could help modernize some of the processes and the ways of working.

There is also a better sense of what government should do more of: take ownership of more products in-house, led by smart technologists who actually know how to build, and constantly iterate and improve. Jen Pahlka has labeled a product approach. Direct File represents what is the most visible and ambitious version of the product approach I’ve seen in American government.

Merici Vinton, who helped to lead the Direct File project wrote:

The approach we took to Direct File is the blueprint for future government delivery — multidisciplinary teams, focused on a clear user outcome, delivering in an iterative manner — and while our project might be dead now, you can’t kill a good idea whose time has come.

Private contractors will still play a role in the product approach, but they cannot be in the driver’s seat. For public-facing products in particular, the building needs to incorporate human centered design — constantly looking at user experience, finding and removing pain points. One Treasury employee contrasted working on another project with contractors where:

if you wanted to improve the product, it was this awful game of telephone where you're like totally dependent on, on like a chain of telephone to these contractors who like, may or may not be an A team….They're not really that accountable to you or to the ultimate mission. And it's also just like a kludgy way of operating. Conversely, like the promise of Direct File was to make it a center of gravity not just for itself, but for modernizing the IRS more broadly and for pulling in other parts of the tax system that also needed to be brought into the tax 21st century.

The hollowing out of state capacity

This will be the first tax season after more than 28,000 IRS employees were fired or left, more than 1 in 4 employees, and wants to cut 4,000 staff. Secretary Yellen pointed to the negative effect on IRS capacity: “There is no more money for the IRS. DOGE came in and ruined it.”

After adjusting for inflation, the IRS budget is 40% below its 2010 level, even as it has taken on more responsibilities and serves more taxpayers. And to be clear, this does not save America money: it blows a hole in the deficit. According to the Yale Budget Lab:

The IRS reductions from funding and layoffs have likely resulted in about $861 billion in decreased revenue. The layoffs from actions like DOGE alone have likely resulted in $597.8 billion in decreased revenue over the 2026-2035, and the clawback in IRS funding of $20 billion has likely resulted in $262.8 billion in decreased revenue over the same period.

It remains to be seen how much of an effect this downsizing will have on tax services, but when you are asking HR staff to answer customer service hotlines, the expectations for world-class service will be low.

The cuts had had a very clear effect on Direct File team, which was dismantled, resulting in many skilled technologists leaving government. In all, IRS eliminated about 40% of its tech staff, mostly IRS tech employees who worked outside of Direct File.

One former IRS employee told me that the DOGE official who killed Direct File expressed surprise that people were quitting, asking the team “Who’s gonna do the work?” Because the Direct File team were generally new IRS hires, they were probationary, and so had little job protections, with a widespread expectation that they would be fired as part of broader elimination of probationary employees pursued by DOGE. “No no one ever bothered to talk to them or reverse that sentiment.”

Many who worked on Direct File drew a sharp contrast with DOGE and their approach to building tech products. One point of distinction was DOGE’s seeming disinterest in public interest goals and of the public itself: “if you do not think government has a responsibility to serve people, I think it draws into question how good are you going to be at making government work better for people if you just don’t believe in that underlying principle”

Towards an attitude of wisdom about government reform

Every year I have my students read Karl Weick on sensemaking. One thing I hope sticks with them is what he calls an “attitude of wisdom”: the ability to combine knowledge and doubt, humility with a desire to challenge the status quo. Wisdom involves experience, but also questioning and self-questioning, a willingness to be wrong and adapt.

I thought about the attitude of wisdom when I interviewed Direct File team members. They had to understand the intricacies of the tax code as well as build a good digital product, and noted the tendency of DOGE to underestimate the challenges of managing public services. One put it like this:

I think that anyone who comes into government and wants to make change goes through a period of, like, understanding just how difficult it is to accomplish and just how just what the scope and scale of complexity that you’re dealing with…someone should patent a law that states that the easier a project looks in government, the harder it is to do in reality. The justification being: if it looks really easy and common sense, it means that 20 people have already thought about it and tried it before you, and they’ve all failed because it’s still around…They don’t know what skills are necessary to do this work because I don’t think they understand the complexity of the work. And meanwhile are creating an environment where the people who do know how to do that work either are being forced out by being fired or are not interested in sticking around.

The leads me to propose a paradox of government reform: the simpler a potential change appears, the more likely that it has not been implemented because it features deceptive complexity that others have tried and failed to resolve.

It is also the case that there are often lots of well-pickled myths that contribute to government sclerosis: we can’t do something because we assume we can’t do something, or because one overly-cautious lawyer said we can’t do it years ago.

It can be both true that outsiders underestimate the degree of complexity in changing government, and that insiders overestimate the barriers to change. Sometimes people in government resist change because of an attachment to the status quo, or because the have not tested alternatives. And sometimes they waste a lot of time not learning from failed reforms of the past, repeating the same mistakes.

And this is where an attitude of wisdom would help is knowing the difference between these two scenarios. It is knowing to ask what were the original statutes or rules that generate the constraint — maybe they actually don’t do so. And sometimes it is trusting employees that constraints are real. How do you know the difference?

In the case of the Direct File employees, I don’t think anyone could reasonably accuse them of being attached to the status quo. This is why their testimony of the need to master the complexity of their policy setting feels so striking and credible to me. Maybe DOGE will get around to building great things, but they haven’t so far, even as they pushed out the people that were. They were impatient, and assumed they knew better.

Can we have nice things?

Yes, we can have nice things. We can have a government that builds cool stuff that helps the public. The thing I’ve heard again and again from tech workers in government is the tech is the easy part. Changing the systems in which they work — including procurement and personnel systems — is the bigger lift.

Fixing a political economy where politically powerful actors want them to fail may feel impossible, but the Direct File team gave those actors fewer fig leaves to hide behind. They made the barriers to public interest capacity clearer. One described their hope that Direct File would return in the future:

my hope is that we work to raise the expectations of the political class that make those decisions and that we are fully empowered to do that and that there’s enough of a mass of people to do it as well… that we have really completely reimagined organizations that are just like able to deliver with ambition and scale of Direct File.

A relatively simple account of Direct File is that one presidential administration dedicated political resources and commitment to building a state capacity in tax reporting, while another opposed it. A more sophisticated account considers the role of private interests. The two arguments can be combined: political commitment to public interest capacity was needed to overturn private interest opposition, but the capacities produced are vulnerable to removal if political support is not broad-based.

Long-term private opposition to a public tax option was able to take advantage of a shift to an administration run by officials who doubted the value of public interest projects, and ended the Direct File after consulting with private sector opponents and with little attention to feedback from the public. The Direct File team built a great public interest tool. Then the CEO of an Andreessen-Horowitz-backed start-up, his hedge fund boss, and the private companies that want to kill competition for their products got together and ended it.

Back to the Free File Future

What happens now? Democrats are pushing to reinstate Direct File, and canny Presidential hopefuls might float it in their campaigns to offer a “get shit done” contrast to Trump. The contrast is easy to make, because the Trump administration’s best idea is to go back to the future.

Last year the One Big Beautiful Bill Act directed IRS to look for private partnerships to provide free filing. Folks, we tried this already. It is called the Free File partnership. It has worked poorly for the very good reason that Intuit and other companies ensured that the product was lousy, and used it to redirect customers to their fee-charging products. Indeed, Intuit repaid $141 million to low-income customers for charging them for services advertised as free, and left the Free File partnership after IRS dropped a non-compete clause that blocked Direct File.

Secretary Yellen pointed to the problems of working with the private tax industry as a motivation for Direct File: “And I think the experience maybe we’d had in the past with Intuit or others developing software that the IRS then didn’t control was a negative past experience that suggested doing it in house would be a better way to go.”

And this is the Trump administration solution? Because in America we can’t do the obvious, proven thing to fix problems. We must replay past mistakes, at the public expense, because it is in someone’s private interest to do so.

P.S. Viktor Orban lost his bid to hold power in Hungary yesterday in such a landslide than any attempt to contest the election was not feasible. I wrote in 2024 about how Orban had eroded democracy in Hungary. Pour one out for the right wing intellectuals who Orban paid to boost his political standing and bring these ideas to America: they just lost a generous patron.