Why the Paperwork Reduction Act needs to go, Part 1

If we have fewer federal workers, we'd better choose what's most important for them to do. A law that keeps us locked into nineties era paperwork should be first on the chopping block.

I don’t know if you noticed, but it’s PRA Week! That’s right, like Shark Week, but the drama comes from obscure statutes and their unintended consequences instead of rows of razor sharp teeth. Marina Nitze’s unsparing takedown of the Paperwork Reduction Act and Kevin Hawickhorst’s excellent history of “forms control,” the Truman era paperwork reduction strategy that actually worked, both came out yesterday. My Niskanen colleague Alexander Mechanick is also publishing his analysis and recommendations for PRA reform today, which are valuable but sound modest in comparison to Marina’s view that the statute should be “nuked from space.”¹

I tend towards Marina’s position, but I’ll get to why over the course of a few posts. There’s a surprising amount to say about the Paperwork Reduction Act because it's a case study in a wide variety of common dysfunctions. For one, it’s easy to see the “cascade of rigidity” in action here; the way agencies interpret the statute is far more onerous than the law or even the office responsible for compliance intend. For another, it acts on what’s essentially an anachronism, like many statutes that have simply been overtaken by events. But ultimately, its theory of change is just wrong: it operates at a chokepoint too far downstream to be useful. It’s trying to improve something through compliance and control that is mostly improved through the creation of capacity where the work actually happens.

In a nutshell, the information collection provisions of the PRA (the law has a bunch of largely unrelated stuff in it) state that before federal government agencies can collect information from 10 or more people, agency officials must post their collection plans for public comment on the Federal Register twice (once with a 60 day comment period and once with a 30 day comment period) and get approval from the Office of Information and Regulatory Affairs in the White House. If you added the 60 and 30 to get 90, you’re already starting to see the problem here: you’re at a three month delay before you even account for the time it takes an agency to put together a package for review, run the all the traps internally, bring it to OIRA, review and respond to the comments from the Federal Register (twice) and go through the OIRA process. In practice, this has often been described as a minimum of nine months and often more, (see Marina’s examples of much longer) though recent years have seen more effective use of “fast track processes” that can bring that time down significantly in certain cases. It’s not just the time that goes by while all this is happening, time during which the information in question isn’t being collected, it’s also the effort expended by a wide variety of stakeholders. One back of the envelope estimate landed at 375 person hours for a new collection, somewhat less for an existing information collection, each of which must be reviewed every three years under the current statute.

Technically, OIRA tracks over 10,000 collections of information collection, so to speak. In other words, each of those 10,000 can include any number of different forms the federal government requires someone to fill out. (On a practical basis, an “information collection” is very often a form, like one you might fill out to file taxes or apply for benefits, and I probably over-index on that use case, but they can also be surveys, questionnaires, and the like, sometimes used for research or regulatory compliance purposes.) Since no one knows the average number of forms or surveys in each collection of information collection, we don’t actually know how many distinct forms OIRA tracks for initial and ongoing review and approval, so it’s difficult to estimate the number of hours a year that public servants spend complying with this law, or the total cost to the federal budget. But it’s safe to stay that for a law most people have never heard of, they’re both pretty big numbers.

I think I can convince you that even before the cuts to the federal workforce started, the PRA cost us far more (not just in dollars or hours, but in effectiveness) than it bought us. But if you read no further on this rather niche subject, I hope you take this away: we can no longer afford to focus our limited human resources on this kind of process theater. Pretty much every agency has lost staff, with even bigger reductions in force coming. By the fall, there’ll be precious few people left in many agencies. Like it or not, we have to make choices about what the federal workforce will do. We should choose the activities that most meaningfully support the mission. If we choose to retain complex and time-consuming procedures like this one, other critical work to serve the public will simply not happen. We will all feel it. We will all regret it.

But that is an observation that is very much of the moment. Taking that action would be a good thing in my view, but it’s also helpful to pull back and take a longer view of how we got here and what that teaches us about the traps we fall into that erode state capacity over time. On PRA, we’ve gotten stuck in an unproductive conversation about how narrowly or widely we throttle an awkwardly conceived bottleneck without recognizing it that never served its intended purpose in the first place. Debating how many comment periods should be required and how frequently forms should be reviewed traps us in our current unhelpful constraints. The work at hand is reconsidering how government should interface with business and the public in the digital age. To understand the nature of that work, it helps to look backwards, forwards, and outside the federal government.

The PRA does not – and never has — reduced paperwork

The PRA came about primarily as a response to concerns from the business community about what we now call the “time tax” of complying with government’s attempts to collect information, often driven by regulation. Alex’s paper offers a thorough and enlightening history of the issue dating back to FDR’s Administration, when effective administration of New Deal programs required data, and requests for information from industry increased. More recently, the burden on individuals, especially low-income communities, has been in the spotlight, as writers like Annie Lowrey and academics like Don Moynihan and Pam Herd have documented the enormous burden of applying for and retaining public benefits.

The problems are old but evergreen: it’s time-consuming to provide the government information about your business or family, and it’s worse when a) there seem to be too many questions, asked too frequently, b) the questions seem intrusive and/or the answers are not easy to provide, possibly requiring the individual or business to find or begin collecting information they had not previously tracked, (for instance, data about a business’s workforce demographics) and/or c) various government agencies ask for the same information, sometimes in different ways, often using different definitions. These problems are exacerbated by federalism, in the sense that just as the public doesn’t see requests for information from disparate federal agencies as coming from separate entities — “shouldn’t they know this already?” — it’s also hard to explain why state and local government, with each of their respective separate departments and agencies, might be asking for the same information yet again. The problem the PRA tries to solve is a real one. But it doesn’t solve it.

PRA’s theory of the case seems to be two-fold. One, OIRA’s review process would allow the White House to tell agencies either to get certain data/information from one of their peer agencies rather than asking the business or individual again, or not to ask for that information at all, reducing the burden of government forms and surveys generally. And two, that the friction required to get approval would make bureaucrats think twice before deciding to collect information at all, thereby avoiding the creation of new forms altogether in addition to streamlining existing forms. If either of those had worked, there would have been a decrease in collection of information starting after the passage of the PRA.

The evidence isn’t perfect, but it strongly suggests otherwise. Stuart Shapiro’s analysis shows 4176 information collection requests going through OIRA in 1986. In 2011, that number was up to 4805.² The PRA was substantively strengthened in 1995 and OIRA began calculating not just the number of information collections, but also the “burden hours,” meaning the number of forms multiplied by the number of people who would fill them out multiplied by the average number of hours it would take to do so. In 1997, that number was 6,970M. By 2010 it was up to 8,783M. Today it is over 10.5B hours.

We might have expected to see a decrease because of the controls in place designed to reduce burden, but there also should have been a decrease because of what happened over the span of those years; the digital revolution. In 1986, not only would someone filling out a government form have done so on paper, but the retention of that data might also have been on paper. Today, both the form and the back end storage are more likely to be digital. This represents a significant opportunity for burden reduction in that data can be both inferred and reused. In the eighties, to make a bank deposit, you filled out a form with your name, address, and account number at minimum. Today you open up the app on your phone and need none of that. In the eighties, you called a dispatch unit and gave them your origin and destination when you needed a taxicab; today, the app knows where you are, who you are, and how you will pay. These affordances have reduced the burden on users of private sector services. While government services have certainly evolved, we don’t see the same kind of dramatic reduction in burden in the public sector— most government forms still start with name (individual or business), address, date of birth, etc.

Our forms still look like they’re from the eighties and nineties in part because the law that governs them was written then (PRA was passed in 1980 and revised in 1995.) It was designed to deal with literal paperwork, so that's largely what we still have. Even if our forms are now digital (some are, many aren’t), they haven’t evolved the way depositing a check or calling a cab has. They are digital versions of the paper that preceded them, because that’s what we have an approval process for.

Even modest attempts to take advantage of what the digital world offers fit poorly in a framework from the eighties and nineties. Marina Nitze (in a previous article) describes hitting up against this disconnect when she was trying to make a better website for veterans in her role as CTO of the Department of Veterans Affairs.

As part of the aforementioned new website, I wanted to have one form "wizard" that would allow a veteran to enter their information once, and automatically apply for all the benefits for which they were eligible. OIRA told me that to do this, I would first have to submit every possible permutation of this wizard for approval—a request I would have found delicious to comply with, were there enough trucks on the planet to deliver that amount of paper.

This isn’t primarily a problem of OIRA staff being rigid or pedantic. This is staff doing the job the law tells them to do. In doing their jobs, they’ve not only stopped Marina from implementing her wizard, they’ve also often stopped other practices that would have reduced burden.

The private and social sectors routinely conduct what’s known as “user research” with representative customers of forms and processes — these practices can include unstructured interviews to understand the users’ constraints, circumstances, and needs, observations of representative users as they try to fill out a prototype of a form, and other ways to understand what in the proposed information collection makes sense to its intended users and what does not. But, until recently, this research itself has been typically subject to PRA review as a matter of practice— in other words, in order to conduct the research needed to reduce the burden of a form, an agency staffer would need to go through a very time consuming nine month approval process, effectively making that research somewhere between impractical and impossible. Thus, most forms in government were never tested with users, and those responsible for creating them had little understanding of the barriers users encountered, increasing the burden. This is just one way a law designed to reduce the burden on the public can in practice increase it.

Recent guidance from OIRA should go a long way to enabling user research, but this guidance will take a while to propagate throughout agencies, and it of course still leaves the months-long review process for forms intact, slowing agencies down and costing them ever-more precious resources. We still need to grapple with a law designed to control literal paperwork, operating in a world where paperwork is an anachronism.

If the goal is to make things easier on people and businesses dealing with government, tuning the dials of our current anachronistic process won’t do much. A more effective intervention would have two major characteristics: it would operate far upstream from the chokepoint of reviewing a form, and it would shift the incentives for agencies from compliance with an outdated law and process to actually improving research and service design outcomes.

Conservative and liberal uses of PRA

Before we look at the proper levers for reducing burden in the digital age, let me first address a likely criticism of my analysis so far. While the PRA has failed to achieve its goal of burden reduction because burden hours have gone up over time, not down, it is also true that during that time we have asked the federal government to do more and more. Of course agencies have to collect more information if you give them new programs to administer.

But for “drown it in the bathwater” conservatives, that’s exactly the point. Project 2025 loves the PRA and calls for “robustly using its authorities” because its authors see the law as a tool for stopping government from doing more than it should. It is, to borrow the subtitle from Don Moynihan and Pam Herd’s excellent book Administrative Burden, policymaking by other means. Sure, Congress may have authorized a new program, but the PRA creates friction in implementing it. It is meant to create a chokepoint a conservative White House can use to thwart agencies from fulfilling those new mandates. And, of course, it doesn’t stop government from doing more. In fact, it makes government do more — by making it jump through more hoops.

I’m sympathetic to the cause of limiting government’s scope, as evidenced by my picking on the USDA Recipe Finder yesterday. My paper with Andrew Greenway identifies “trying to do too much” as one of the core dysfunctions of government. If we have limited resources, we should focus them on what matters most. But if a program should be cut or reduced in scope, the time to make that decision is not when it’s getting approval for an information collection. It’s too late to kill or rescope it, but now it’s a huge lift. You needed to have that debate upstream.

Conservatives want the PRA to be a tool of small government. Recent Democratic administrations have seen the PRA more as a tool to improve customer experience. The Biden administration went all in on this, issuing an executive order on CX and engaging OIRA deeply in the roll out of it. Talents like Amira Boland put enormous muscle behind fast track processes to minimize the burden on agencies. (To be fair, she started doing this under the first Trump administration.) Alex recounts the heroic and expert work of staffers like Kyle Gardiner, whovused the review process to push the Social Security Administration to make application and recertification processes easier. One of the brightest spots of user-centered policymaking happened in OIRA in 2024 when the team there, led by Sam Berger, decided to make it clearer to agencies that user testing did not require PRA clearance (since unstructured, observational research is exempt, as Erie Meyer has long pointed out) and tested its guidance clarifying this with the agency staff who would need to interpret it. As usability testing often shows, what you thought you’ve said isn’t what people have heard, and the team revised their guidance several times through these tests to ensure that agency service designers would truly be allowed to conduct their own research and testing. These efforts and others made the PRA clearance process less of a barrier and more of a value add.

But it’s still a barrier, and you’re still operating at a chokepoint too far downstream to be meaningful. Say you’re an OIRA desk officer and you’re looking at a form and deciding whether to approve it or not. You could say “well, we don’t think you should be asking this question at all. If you need that information you should try to get it from another agency that already has it.” But at that point the form may have been through many months of design, many months of packaging up for approval, and maybe even many months in the approval process. Negotiating a data sharing agreement with another agency in order to be able to remove that question from the form (or pre-fill it, as appropriate) could take years, and as a practical matter, it’s hard to send it back for that. You can tell them that the next time they come for approval, they will need to have gotten that data elsewhere, but that’s a process with multiple failure points over a long period of time.

You could say “Hey, we don’t think you should ask that question that way” but you’re very unlikely to be a representative user. Maybe it’s a form for farmers getting USDA loans, maybe it’s for veterans applying for benefits, maybe it’s for local public health officials submitting data to the CDC — the whole point of user research is that what makes sense to someone who creating (or approving) a form often doesn’t makes sense to someone filling out that form. If the form has been tested with representative users, the OIRA desk officer’s opinion about how to word questions is a lot less relevant than the results of the user testing. If the form hasn’t been tested, then it’s probably pretty confusing. Either way, a meaningful check at this point in the process is less about what's asked in the form and more about what practices were used to make those decisions. And either way, because of how long all this takes, it’s too late. The Tappet Brothers used to say at the end of their hilarious radio show Car Talk, "if you want an audiotape of this show, you should have pressed record an hour ago.” If you wanted a lower burden form, you should have done user research nine months ago.

This is part of what Kevin Hawickhorst is getting at in his excellent piece on how federal government used to make forms better. The Truman administration instituted something called Forms Control, which sounds a lot like an information collection review chokepoint, but isn’t. For one thing, there was no chokepoint – it was about enablement and capacity building within agencies, not centralized control. There was a design language and proposed processes. It stressed tools over rules, but to the extent there were rules, they were self-policed. There was no chokepoint downstream because they had invested in competencies upstream, where the people doing the work were closer to the problems that needed solving.

It turns out that works. Apparently we used to know this. We’ve forgotten, but the rest of the world picked up that strategy and ran with it. Tomorrow I’ll cover paperwork reduction efforts (as it were) in our peer countries and what the US should learn from them.

Stand by!

1. https://reason.com/2025/02/13/i-tried-to-fix-government-tech-for-years-im-fed-up/
When Marina calls for repealing the PRA, she really means the information collection provisions of the law. Its not entirely clear if anyone would notice if the rest of it were gone too, but there is a lot in the PRA that has nothing to do with OIRA’s review process.

2. https://www.sciencedirect.com/science/article/abs/pii/S0740624X13000087