More (Brief) Thoughts on DOGE

What have we learned in the last 3 months?

Last week I was traveling,¹ this week I am attending a wedding, and the past month has been extremely busy, and as a result I have not been able to finish an essay on where DOGE has gone. In lieu of that imagined brilliant essay, here are some scattered thoughts on the state of play: where DOGE stands now, what I got right and wrong, and what we’ve learned.

Context

The thoughts on DOGE in early March I wrote up garnered some attention. But it’s been 3.5 months, and everything has changed. Elon’s out, having called the president a pedophile and threatened to start a new political party. MAGA hates Elon, and multiple DOGE team members have left, with at least one doing a media tour and others going off the grid.

For a printable transcript of this piece, click here:

What I got wrong

I tried to be as clear-eyed as possible when covering DOGE earlier this year, and I’m happy with how the piece held up in a murky information environment. I also tried to be explicit about things I’d already gotten wrong over the winter, pre-DOGE’s formal beginning. But there are a couple things I missed.

For instance, many people (including many of you) predicted that Elon and Trump would have a falling out in the near future. I did not think that would happen as early as it did. I paid a lot of attention to the genuinely close personal relationship the two of them seemed to have (and despite the catastrophic falling out, it’s still more of a love-hate relationship than an outright enemies thing). I paid less attention to my own gut sense of just how grating Elon appears to be personally, and to how quickly he was burning through political capital (and burning through basic relational good will) within the inner circles of the admin.²

I also rather overrated the level of talent entering DOGE. Not by a huge amount — there were some genuinely impressive technical minds on the team — but there were not all that many of them. I reasoned from too few examples here. Also, plenty of them were not asked to put those talents to full use, or directed them toward the wrong goals. Some of the quite intelligent DOGE team members were/are also true believers a la Musk that slashing rampant fraud in the federal budget could solve the national debt, which made them straightforwardly worse at using their skills to identify and resolve real problems in government inefficiency. The recurring inability to read federal contracts, and the inability to get a handle on the real amount of phone fraud at Social Security, both come to mind here.

I could have said something about how the fun of running DOGE played into all this. The critique that it was a billionaire’s plaything was basically right, with an emphasis on play. More on this in a second, but you can make a coherent argument that Elon paid attention to DOGE for as long as it was fun, and stopped at precisely the moment it stopped being fun.

What I got right

• The Vivek version of DOGE would have been better both politically and practically.

• Elon was and is genuinely fixated on the national debt. This claim received a lot of pushback earlier this year,³ but it was always true, and is now more obviously true. Yes, you can’t solve the national deficit by ending fraud in Social Security. Yes, Elon genuinely believed that you could, that there was close to $1 trillion in fraud to be prevented in Social Security, and that large chunks of discretionary federal spending is fraud too. Yes, that belief makes no sense to anyone who pays attention to how the federal government functions, yes Elon held it anyway, yes, Elon indulges in motivated reasoning often and in public.

• Elon’s tendency to fixate on legible metrics is a big part of his success at Tesla and SpaceX, and drove a lot of the bad outcomes for DOGE. Again, this seemed obvious as an explanation, and I think is only more obvious now.

• As is Elon’s susceptibility to information feedback loops on Twitter. Most obviously, the catastrophic decisions to break USAID emerged out of the Twitter feedback loop.⁴

Easy mode

Political institutions follow the path of least resistance, unless their leaders work actively to avoid that flow. Why did DOGE end up cutting the contracts of the people watering office plants, and not pursuing the tens of billions of dollars of fraud mainstream observers acknowledge exists within Medicare Advantage? Because the office plant thing is easy, and Medicare Advantage is hard. Elon’s insistence on legible metrics, on “What did you get done this week” speeds, and on raw dollar amounts saved on paper led to the disproportionate cutting of cheap, vulnerable, high-value work in the federal government.

Coordination matters

There’s been excellent mainstream media coverage of DOGE, so someone may be able to confirm or rebut this, but my understanding is that there was never any central channel for all DOGE team members to communicate. It was all multiple cells up to Steve Davis, who reported directly to Elon. It’s also my understanding that DOGE completely stopped doing all-hands meetings in March. As a result, the amount of information sharing across DOGE itself was quite limited. This seems like an unforced error borne out of worries about loyalty and leaking. It also suggests that you can fail to capitalize on the efficiency gains of getting a live data feed of how the federal government is spending money, by not actually using that data.

A timeline

• March 6th: Trump says Elon won’t get to decide who gets fired.

• Also sometime in March: Regular all-hands meetings with Elon end, and other reporting says Elon pulls back from DOGE/his interest begins to wane.

• Sometime in early April, Scott Bessent and Elon had their physical fight in West Wing, says Steve Bannon.

• Elon’s tweets about DOGE slow down and his tweets about SpaceX increase, and then on

• May 30th, before and after Trump gives Elon his public Oval Office sendoff, head of PPO Sergio Gor pushes Trump to scrap the nomination of Jared Isaacman, an Elon ally, to lead NASA.⁵

• Elon crashes out publicly, calling Trump a pedophile, etc.

Did the RIFs work?

No doubt the administration (largely through DOGE) has cut a lot of people from jobs in the federal government. How many people, though, is hard to quantify, both because there are active lawsuits over some of those firings, and because across the federal government, agencies are having to rehire employees to achieve administration priorities. For example:

• Late May: The IRS rehires thousands of employees, purportedly fired for “performance issues.”

• June 2: After losing more than 500 employees to cuts and early retirement incentives, “The National Weather Service has secured permission to hire about 125 new meteorologists and other specialists… The weather service is also still moving forward with a stopgap measure of transferring meteorologists from well-staffed offices to ones that are down multiple meteorologists, radar technicians and other specialists. NOAA announced last month it was seeking 155 transfers to fill these ‘critical’ positions.”

• June 6: The FDA rehires employees. At USDA, the Forest Service asks staff who took the “hard fork” to come back, because it’s fire season.

• June 11: HHS tells laid off employees (around 500) that they are un-RIFed. "That notice is hereby revoked.”

The defense of this state of affairs would look something like, “Elon engineering principles! If you haven’t cut too much, you haven’t cut enough!” This principle makes plenty of sense in engineering, and less sense when you’re trying to staff the Forest Service for fire season.

Did DOGE streamline procedures?

In some places, yes — see the early retirement work led by Joe Gebbia for an interesting example. And in other places, it’s added basically reasonable procedure to limit fraud. For example, “The Dept. of Education, starting next week, has now instituted mandatory identity checks for all first-year students receiving financial aid.” DOGE cited international crime rings defrauding the Department of Education as the reason to institute this extra layer of process. I was curious whether those crime rings really exist, so I asked Linda Miller, who focused on fraud prevention as Assistant Director at the Government Accountability Office. I’ll footnote her whole answer, but her takeaway: yes, this is a totally reasonable thing for DOGE to focus on, and will probably save the taxpayer money.⁶

But to a large degree, DOGE has added meaningless bureaucratic procedure. For instance, DOGE put a $1 spending limit on government-issued credit cards, citing risks of fraud. As a result, hundreds of thousands of minor processes are held up in the government, as employees wait for phone bills, paper, pens, and printer toner.

• “Things like web subscriptions, payments for routine services have become a huge hassle. An hour or two of staff time would now go into what would have been a quick internet purchase for something like buying a $10 box of Ziploc bags for collecting samples — highly efficient, isn’t it?”

You can see a similar dynamic outside DOGE, like at the Department of Commerce, where Secretary Lutnick reportedly has massive amounts of paperwork to churn through after instituting a rule that he or a high-level political appointee must sign off personally on all spending over $100,000. I’ll have more on this topic in the future, but this kind of problem is historically common in attempts to get more direct control over the executive branch (Nixon faced similar paperwork burdens, for example).⁷ Clamping down on perceived political independence at lower ranks of a department can push operational tasks up to higher ranks, reducing the ability of political appointees themselves to make political decisions.

What’s next for Statecraft+DOGE

• I would love DOGE team members, current and former, to come on Statecraft, and I’ve invited multiple. So far, none of them have bitten. But there’s a lot to talk about. That invitation remains open!

• There is plenty of good government reform work happening in the federal government right now, alongside some boneheaded moves. As ever, my goal will be to try and give accurate accounts of both, and to let readers decide how they weigh against each other. Statecraft will continue to “platform” people readers may find objectionable (and I don’t really believe it makes sense to complain about very powerful people one dislikes being “platformed” by us or any other outlet.)

• I’m working on a couple essays about what we can learn from past government reform efforts. Hopefully will have more on that topic in the coming months.

1. While on vacation in upstate NY, I got to tackle an advance copy of Dan Wang’s new Breakneck: China’s Quest to Engineer the Future. It’s great. We’ll try to get him on Statecraft this fall.

2. In conversations with people close to DOGE, I’ve pretty consistently heard them articulate a particular theory: Elon had to move fast to get things done, and the only way he could do that was by spending lots of political capital very quickly, and he burned out before he could defeat his enemies.

My view is that if you’re going to think in the metaphor of political capital, then you should also think about it as something you can accrue. Doing popular things increases the amount of political capital you have to work with. So does helping allies. Success gives you more political capital to work with, not less.

3. Ezra Klein, to me, on Elon’s motivations: “I’m very skeptical of this cutting-the-debt theory.”

4. I think gutting USAID was a bad decision, and I’ve been consistent on that point. We have a great episode upcoming with former Chief Economist of USAID Dean Karlan talking about it. But some critics saw equivocation and “victim-blaming” in lines like this from my March piece: “There are good things USAID funds, and wasteful things it funds, and good governance means being able to distinguish between the two.” To which I would say, separate from the issues with USAID documented in that piece, the Justice Department recently “secured four guilty pleas for bribery to get over $550 million in special race-based USAID contracts.” It’s completely consistent to think that USAID was ripe for reform, and that DOGE made a great moral mistake in freezing humanitarian aid. I trust readers to hold more than one belief at a time.

5. Canning Jared Isaacman seems to have been motivated largely by the spite of Sergio Gor, who runs the Presidential Personnel Office and feuded with Elon. I think it’s a huge mistake to pull Isaacman from contention, and it will be hard to find someone nearly as good as him. If you were a tech-aligned space innovator, would you accept the nomination now?

6. Linda Miller:

• Yes, this is accurate. Domestic and International crime rings--professional fraudsters--have been exploiting the relatively lax identity verification processes of both the Department of Education and many smaller colleges and universities to steal hundreds of millions through this "ghost student" fraud scheme for several years now. It seems higher ed and DOE caught on about a year or so ago.

I'd say a couple things:

• Criminals have a near infinite supply of personal identity data that they can and do monetize in a startling variety of fraud schemes across different industries, with government and smaller organizations presenting an easier target than larger private companies who generally have more sophisticated systems to check for stolen identities

• The 20% statistic DOGE is citing comes from a California report. California doesn't require SSNs of students applying to its state colleges, making this scheme even easier.

• The rise of remote learning since the pandemic created the conditions for this scheme to flourish. It's possible to enroll in college and never have to show up anywhere in person, even as a legitimate student, so the ghost student scheme has been easy to hide.

• Organized crime--both domestic and international---has moved online. Street gangs are now cybercriminals. It's just too easy to cash in on the ocean of identity data available for sale as a result of almost daily data breaches. AI tools have enabled fraud to scale with alarming speed and ease.

• By and large the government is oblivious to this threat. Agency leaders I speak with regularly tell me fraud is not a problem at their agencies, even those doling out billions in grants and loans.

• These crime rings are very good at concealing their work. They use a huge network of groomed mules to launder their stolen funds allowing them to operate undetected.

7. There’s a pendulum dynamic that’s happened in government reform: If you centralize control over spending, it creates bottlenecks on small processes, so reformers try to open up spending (we covered an attempt at this recently). Over time, decentralization of spending leads to multiple duplicative and genuinely wasteful processes, which leads to the push to centralize again, etc. I want to spend some time on this issue in future and would be curious to read/hear more from any readers who have observed this dynamic.