The Case for a "Department of Government Efficiency"

There ought to be a federal department exclusively focused on making government work better, that is, government efficiency.

It makes sense. After all, there are many ways that government could be more efficient. For example:

• Government-funded scientists say that they spend 44% of their research time on paperwork and bureaucratic requirements. That’s just an average. Anecdotally, I’ve heard from scientists who spend up to 70% of their time on bureaucracy. We should drastically reduce the administrative burden on scientists.

• The Paperwork Reduction Act paradoxically results in endless paperwork for government employees, and it’s not clear that anyone else benefits at all. We should repeal its information collection requirements, or at least raise the threshold for review and approval (i.e., current law requires bureaucratic review for any action that affects 10 people or more, which is an insanely low threshold for a country of well over 300 million people).

• The National Environment Policy Act (NEPA) creates an enormous and unnecessary paperwork burden for new infrastructure, and ironically has been used to block clean energy.

• Federal procurement is broken, with far too many inefficient rules that waste time and money, and that result in the government buying over-engineered products that don’t even work. We need extensive reform of the Federal Acquisition Regulation (FAR).

Many such cases.

We should have an official effort to address these issues (and much more). We could even call it a “Department of Government Efficiency,” or DOGE. Such a cross-government agency could have a high impact on government efficiency, by providing a way to break out of the usual bureaucratic chains of command.

How should this new DOGE do its work?

First, they should spend a great deal of time developing an in-depth understanding of how each government agency works, how its data systems are structured, what regulations affect its work, and more.

Here’s an example of why it’s important to take your time. I was on the board of the Houston Education Research Consortium (HERC) for many years—a collaboration between Rice University scholars and the Houston school district (one of the largest in the country). For more than a year, HERC had to figure out how to process the Houston data. It was frustrating. I repeatedly heard stories like this:

“There was this data field titled something like XP518VC, and we would ask what that data field was tracking, what it meant, etc. We’d get the response that Mary down the hall used to know what that data field meant, but she retired two years ago.”

Or:

“We saw that in one year, the number of Native Americans skyrocketed in Houston. Like, it went up ten-fold. We couldn’t figure out where all these Native American students suddenly came from. Then we noticed that the number of Indian Americans plummeted in the same year. Turns out that someone got Indian Americans confused with American Indians when entering the data.”

A few points about this:

• Yes, government data systems can be incredibly inefficient and opaque. But:

• It is foolish to cry “fraud” after a day or two looking at a database. It’s much more likely that you don’t understand some key feature of the data, or that there was routine human error, or there was some systemic change in the way data were collected and reported.

  • In some government databases, for example, missing data can be coded as “-99.” Before rushing to announce that an agency is committing fraud by enrolling people with negative ages, it would be better to spend some time with people who know the database inside and out.

• To fix any problems, you can’t expect results overnight. You may well have to spend a year or more working with internal folks just to understand the system, let alone figure out how to fix it.

Second, a new Department of Government Efficiency should be efficient itself by focusing on reforms that would deliver high-value for the time/effort/expenditure. For example, is there a government data system that is 50 years out of date? Is there a government process that takes far more time than it is worth, and could new software speed it along? Are there areas where agency counsel has been far too risk-averse in interpreting the law?

To answer these questions would require a lot of qualitative work (e.g., interviews, surveys, focus groups) with current government workers to figure out what their pain points are. After all, most government workers aren’t complete slackers. In many cases, they took a lower-paid job out of civic pride, and they (more than anyone) want to find ways to make their jobs more efficient and effective.

Just ask them! You’ll hear endless stories and ideas for improving efficiency.

Third, a well-functioning DOGE would learn from the lessons of past government reform—what worked and what didn’t. Similarly, DOGE should work with other government reform groups that have extensive expertise and detailed ideas, but often haven’t had the power to actually carry out reforms. For example, the Council on Government Relations (COGR) has been tracking the burden of bureaucracy on researchers in some detail:

My friends at COGR also recently published a detailed list of regulations, policies, and statutes (in the DOGE-requested format!) that should be streamlined so as to free up scientific researchers to spend more time on research.

A Department of Government Efficiency wouldn’t try to reinvent the wheel. Instead, it would use its power to implement the many reform ideas that have been circulating for years or decades amongst government experts.

As a side note, there are many areas of government where we don’t actually want “efficiency”

Much of government is like a fire insurance policy on your home. It’s much better to “waste” money every year on something that goes unused than to see it put to use because your house burned down.

Consider folks who work on pandemics, or the fire department, or FEMA, or the entire military for that matter. We don’t actually want those people to be busy every day with a pandemic, endless fires and hurricanes, and a world war.

Instead, the best-case scenario is if they never actually have to be deployed at all! Some might think this is “inefficient,” but like insurance, we want those folks hanging around in case they’re needed.

There are many other examples throughout government where the work is cyclical, or in response to outside events, etc. We don’t want all IRS employees to be as busy year-round as they are in tax season around April 15.

To quote strategist Vaughn Tan:

The rhetoric of efficiency is currently all-pervasive. It is almost always borrowed from private-sector ideas of management, propounded by people who made a lot of money in the private sector. But a government is not a private-sector company and cannot be run like one.

An organisation built around efficiency is optimised for a particular configuration of its operating environment. By design, such an organisation is planned with as little slack/redundancy as possible. It is well-run for as long as its operating environment doesn’t change. When the environment changes, it falls apart because it has no slack to fall back on.

Business history is literally full of businesses that disappear when their environments change, to be replaced by new businesses better able to respond to the changed environment. But governments and other institutions with obligations to the public must operate across much longer time horizons. We don’t want governments collapsing and being replaced. . . .

This means that governments and public sector institutions must be built to be ready for their operating environments to change unpredictably over long periods of time. This requires resilience and efficacy. This is why resilience and efficacy should matter more than efficiency in public institutions.

Still, there are some places in government where we should try to increase efficiency. It just takes time and effort to understand where those places are.

Wait a Minute, There Was Already a DOGE?

Just kidding. As it turns out, there is already a Department of Government Efficiency, run by Elon Musk.

But it is now widely viewed as a failure—and I would argue it’s because DOGE didn’t follow any of the advice above.

What did DOGE do instead?

For one thing, DOGE has never had a clear objective. At times, Elon and others would talk as if the purpose of DOGE was to cut the deficit by some astronomical number (even $2 trillion). Anyone who had the slightest familiarity with the federal government knew that was completely impossible, especially given the political commitment not to touch entitlement programs or defense.

At the same time, DOGE’s focus on cutting the deficit has led it to make many destructive choices, such as canceling great education research or firing thousands of government employees for no actual reason. Government salaries make up around 4% of the overall budget—much less than would be typical at a Silicon Valley company, you might note. [After all, our federal government is basically a giant machine for collecting taxes and redistributing the money to defense contractors, elderly people, dialysis patients, and hospitals/doctors that treat elderly people.] It is impossible to make much headway against the deficit no matter how many people you fire, and it just makes government more inefficient to boot!

As well, DOGE has focused on finding (alleged) waste, fraud, and abuse. That’s a reasonable goal, except for a few things:

1. The GAO has already produced a long stream of reports on where to find potential fraud, and DOGE seems to have shown no interest in working with GAO or implementing its many recommendations. Instead, it reinvented the wheel, and even claimed to have uncovered rampant fraud that was completely non-existent (e.g., “tens of millions” of fraudulent Social Security recipients with birth dates over 150 years ago, which is an absurd claim in at least 3 separate ways).¹ I could have missed something, but it’s not clear that DOGE itself has uncovered even one case of actual fraud.

2. There often isn’t that much waste to be found. That’s why DOGE has often been reduced to bragging about trivialities such as cutting a $21,000 executive coaching program or a $56,000 contract to maintain an office’s plants.

3. Most importantly: the goal of minimizing waste, fraud, and abuse is often completely contrary to the goal of government efficiency. Indeed, the most onerous bureaucratic requirements are usually inspired by someone’s desire to eliminate waste, fraud, and abuse.

Occasionally, DOGE at least tried to do something productive to reduce bureaucracy and improve efficiency. In fact, the main (perhaps only) unambiguous success of DOGE to date was to digitize the federal retirement system, which had previously been paper-based.

But those have been rare occasions, and DOGE employees would sometimes admit the truth: “I would say the culture shock is mostly a lot of meetings, not a lot of decisions. But honestly, it’s kind of fine—because the government works. It’s not as inefficient as I was expecting, to be honest. I was hoping for more easy wins.”

So said Sahil Lavingia, founder of Gumroad, who joined DOGE in March 2025 and was deployed to the VA (see also here):

I did not find the federal government to be rife with waste, fraud and abuse. I was expecting some more easy wins. I was hoping for opportunity to cut waste, fraud and abuse. And I do believe that there is a lot of waste. There's minimal amounts of fraud. And abuse, to me, feels relatively nonexistent. And the reason is — I think we have a bias as people coming from the tech industry where we worked at companies, you know, such as Google, Facebook, these companies that have plenty of money, are funded by investors and have lots of people kind of sitting around doing nothing.

Of course, he was immediately fired for telling the truth. As Lavingia said:

And then, my access got revoked pretty shortly after. I didn't get notified. I was basically ghosted and I just got an email notification that my access was no longer valid.

Unfortunately, they did not tell me directly that the reason I was let go was because of my transparency. I don't know if irony is the right word, but I do think that it's maybe, as Elon says, the most entertaining outcome is the most likely, and letting someone go for being transparent in the most maximally transparent organization is a little bit entertaining.

But even the attempt to make the VA more efficient probably didn’t work very well. As ProPublica reported, Lavingia’s code led to many mistakes: “The code, using outdated and inexpensive AI models, produced results with glaring mistakes. For instance, it hallucinated the size of contracts, frequently misreading them and inflating their value. It concluded more than a thousand were each worth $34 million, when in fact some were for as little as $35,000.”

Lavingia himself told ProPublica: “I’m sure mistakes were made. . . . I would never recommend someone run my code and do what it says. It’s like that ‘Office’ episode where Steve Carell drives into the lake because Google Maps says drive into the lake. Do not drive into the lake.”

Amazing! The admission that DOGE has been using hallucinatory LLM models might explain the staggering number of basic errors that DOGE has made, e.g., confusing $8 billion with $8 million, or claiming to have canceled several hundred contracts that were already completed.

Lavingia’s comparison to Silicon Valley companies is also worth a further comment. Companies that are swimming in revenue from ads or VC might indeed have many overpaid slackers, but ironically that isn’t nearly as true in government. Hardly anyone becomes a park ranger or an IRS accountant in the hopes of getting rich.

Indeed, it’s common to talk to people in government who work 60 or 80 hours a week for a salary that is a fraction of what they could earn in the private sector. Indeed, I know a Harvard lawyer who spent nearly two decades working at DOJ with a salary under $200,000, and then finally left for a private sector job making nearly ten times as much.

That happens all the time in the law—government salaries at DOJ and US Attorneys’ offices are much lower than the same people could earn elsewhere. They’re trading current salary for prestige and future opportunities. This happens in many sectors, and it is a weird disconnect that DOGE seemingly thought that all government employees are low-IQ fraudsters. Just not true.

***

Another oddity is that DOGE often just didn’t bother to get its facts straight. It’s not clear why DOGE would so often feel compelled to lie (or to rely on error-ridden LLMs for its public statements), but that’s what it did.

One of many examples: Look at this recent video of Elon and a larger DOGE team being interviewed on Fox News. A DOGE guy claimed that a $4 billion Covid fund was being misspent by schools (such as “renting out stadiums”), and that DOGE created a new requirement: “upload a receipt, that was the only change that was made . . . and upon doing so, nobody drew down money any more.”

Elon added that it could be a fake receipt such as a “picture of your dog,” but “as soon as we asked for anything at all,” suddenly “we don’t need it any more.” Watters then says, “Leasing stadiums, for what?!” Elon says, “for parties, basically.”

As far as I can tell, literally every word in that segment was untrue (except for the word “stadium”).

• There wasn’t a $4 billion Covid fund for K-12 education. It was more like $190 billion.

• Was all the money spent properly? Who knows, probably not, but it looks like all the money was distributed years ago. It’s not clear how DOGE could have had anything to do with stopping new requests (if those even exist).

• If you go to the DOGE Twitter account, it mentions one additional detail: $393k to rent out a major league baseball stadium. If you look into that claim, it turns out that the Santa Ana school district in California, at the height of Covid, decided to have an outdoor graduation ceremony for a bunch of high schools. They publicly announced that it would be at a stadium (where else are you going to host 10,000 people outdoors?), and they literally posted it on YouTube in 2021.

In short, this was a complete non-scandal. A large district at the height of Covid rented an outdoor stadium for a perfectly routine graduation ceremony, not for a party (Elon or his staff should have used some common sense here, given that they evidently couldn’t spend 10 seconds to google anything—since when do school districts have parties at a stadium?). The district did all of this quite publicly, and it was never a secret that needed to be uncovered by DOGE. Moreover, it happened 4 years ago at the height of Covid, and it isn’t an recurring issue where the school district would need to submit receipts to DOGE.

A reasonable heuristic: If a team has legitimate successes worth bragging about, they wouldn’t waste precious minutes on national TV telling obvious lies about non-scandals from several years ago that they didn’t actually uncover and can’t possibly affect.

Another example: Elon and DOGE claimed that Treasury had been processing trillions of payments without an identification code linking the payment to a budget item. I don’t know anything about Treasury payment systems, but experts in the administration are known to have said that "Elon has no f***ing idea what he is talking about."

Even worse, DOGE repeatedly came up with new policies that made government efficiency markedly worse. There’s an executive order now on DOGE approval for payments, contracts, grants, and travel, all of which means extra bureaucracy and inefficiency.

Even worse, DOGE has demanded to approve all credit card expenditures over $1 (I’ve confirmed this with multiple government employees). That policy is absurd. No corporation would make all its executives write a memo asking for permission to spend more than $1 on a coffee or on taking a client out to lunch. The unambiguous result is that government is less efficient. From the Washington Post:

DOGE has also inserted an extra layer of review and approval into grants from NIH and elsewhere at HHS. As Robert Gordon told the Washington Post, “Instead of cutting red tape, they are strangling grantees with it.”

As well, I’ve heard from multiple sources that DOGE has one guy (with no scientific background) who demands to approve every action taken by BARDA, the Biomedical Advanced Research and Development Authority. The result is a striking slowdown of BARDA’s ability to do pretty much anything.

These are all just examples. DOGE’s obsession with the idea that government employees are constantly wasting money has led them to create more bureaucratic practices and layers of review/approval that objectively make government efficiency worse. It’s so bad that government insiders will say things like, “Is the actual purpose of DOGE to cripple the government’s ability to ever do anything?”

Conclusion

With Elon moving on, and with the confirmation of Trump appointees who (quite understandably) don’t want their employees wasting time on DOGE stunts that actually reduce efficiency, we might have an actual Department of Government Efficiency someday.

Let’s hope.

1. First, Social Security already has a policy to automatically stop paying anyone with a birth date over 115 years ago. It takes 5 seconds to find this policy via Google. Second, if there were tens of millions of fraudulent recipients, that would encompass nearly the entire Social Security program, which would mean that the Social Security Administration has been busy sending out tens of millions of checks to people whom they know to be dead while almost no actual elderly people have asked to receive Social Security. Third, anyone who wanted to defraud Social Security likely wouldn’t claim to be 150+ years old.