title: "The Arsonist's Invoice: DOGE at One Year" subtitle: The Department of Government Efficiency spent a year breaking things. The bill is in. author: Rebecca Rowan publication: For the Republic date: 2026-02-14

The Arsonist's Invoice

Here is what the Department of Government Efficiency cost the American taxpayer in its first year of operation.

*One nonpartisan analysis (Partnership for Public Service); excludes litigation costs.

They wanted receipts. Here are the receipts.

Breaking Even Was Never the Point

There's a phrase for what DOGE produced, and it deserves to enter the language: negative returns on destruction. You spend money breaking something. The breaking costs more than the thing was costing you. And now you own both the original expense and the wreckage.

The ledger isn't just underwater. It's spectacularly, humiliatingly underwater. The people who actually knew how government works walked out the door. And a government that can't collect taxes, process retirement claims, or inspect food isn't "efficient." It's incapacitated.

The financial case is already closed. The Cato Institute -- a libertarian think tank that is ideologically built to celebrate government reduction -- found that federal spending rose $248 billion in 2025 even as DOGE executed the largest peacetime workforce cut on record. Two hundred and seventy-one thousand workers gone. Spending? Up. "No noticeable effect on the trajectory of spending," Cato wrote. When the libertarians are telling you your cost-cutting didn't cut costs, the fiscal debate is over.

And the savings claims themselves? WBUR's On Point found that all 13 of the largest items on DOGE's "wall of receipts" were wrong. Not most of them. Not the borderline calls. All thirteen. An ICE contract was claimed at $8 billion in savings; the actual contract was $8 million -- off by a factor of a thousand. A Politico analysis reported by Newsweek verified only $1.4 billion of $52.8 billion in claimed contract savings. Pulitzer Prize-winning reporter David Fahrenthold of the NYT put it flatly: "DOGE didn't cut a dollar of federal spending."

This isn't rounding error. This is fabrication at industrial scale.

Cato's data proves spending wasn't cut. But why was it structurally impossible to cut? Because DOGE attacked the wrong target. And not by accident.

The 8% Chainsaw

Federal employee compensation -- according to Cato's own numbers -- is roughly 8% of total federal spending. The other 92% is Social Security, Medicare, Medicaid, defense procurement, interest on the national debt. Mandatory spending and locked-in commitments that workforce cuts cannot touch. You can't fire your way out of Social Security obligations. DOGE took a chainsaw to the 8% and left the 92% completely alone.

This wasn't a mistake. Cutting the right targets -- entitlements, defense contracts -- would require congressional action and actual political courage. Neither was on offer. So the chainsaw went where it could go, not where it needed to.

A defender might point out that the 8% framing is misleading -- federal employees administer 100% of the budget, not just their salary share. Fair enough. But this actually strengthens the case: if those workers are the load-bearing infrastructure of the entire federal system, then cutting them doesn't reduce costs elsewhere. It just degrades everything they touch. The Cato data confirms it. The service collapse confirms it.

That collapse has a human face. Science/AAAS reports that 10,109 STEM PhDs left federal service in 2025 -- three times the 2024 rate, 14% of the government's entire doctoral workforce. At 14 research agencies, departures outnumber new hires 11 to 1. Those are the climate scientists who tracked hurricanes. The epidemiologists who ran pandemic response systems. The auditors who made sure your tax dollars were spent correctly. 106,636 years of accumulated STEM and health expertise. Gone.

And at the Social Security Administration: 6 million pending cases. Nearly 600,000 retirement claims backlogged -- a 71% increase. Call wait times that peaked at two and a half hours. These are people who paid into a system their entire working lives, now stuck waiting for benefits they already earned, while the agency that's supposed to serve them gets gutted. Real people. Real suffering. And entirely predictable.

The Alibi

I want to be honest about what DOGE actually got right, because pretending nothing worked would be dishonest.

Doge.gov was a genuine transparency win. The Treasury payment tracking fix -- making a previously optional ID field mandatory -- that's real modernization. The CMS Fraud Detection Operations Center suspended payments to 33 fraudulent providers in its first month, including one that had been billing medical equipment for a patient who died 20 years ago. OPM cut contract spending by 50%. These are real things that actually happened, and they're worth acknowledging.

But here's the thing that matters: you didn't need to fire 271,000 people to build a website. The 18F program, USDS, and GSA's Technology Transformation Services had been doing exactly this kind of modernization for a decade -- without mass layoffs. Technology modernization and mass destruction are different projects that got deliberately welded together. Every good thing DOGE accomplished could have been done without the bad.

The question is whether the good things were the purpose, or the alibi.

Defenders love the corporate turnaround comparison: short-term pain, long-term gain. But corporate restructurings are designed to keep your best people and cut redundancy. DOGE did the opposite. The blunt instruments -- hiring freezes, "fork in the road" emails (yes, that was actually what they called them), across-the-board RIFs -- drove out the most experienced, most employable workers first. The PhDs. The senior auditors. The people with options outside government. A restructuring that hemorrhages its best talent and retains its least mobile workers isn't a turnaround. It's adverse selection at scale. It's a liquidation.

Short-term tax receipts held up through April 2025 -- a data point defenders cite, and fairly. But people notice when enforcement weakens, and they adjust. Not in months. Over years. Britain already ran this experiment. HMRC cut staff during austerity and lost 42 billion pounds in uncollected tax. For every pound spent on enforcement, they'd been recovering 18. The UK deficit was worse after the cuts than before. It took years to show up in the numbers. The American data will trace the same curve -- it just hasn't had time yet.

The Yale Budget Lab projects that IRS workforce cuts will cost between $198 billion and $323 billion in lost tax revenue over a decade, depending on how deep the cuts go. Every dollar spent on IRS enforcement yields $5 to $12 in revenue. Cutting IRS staff is, mathematically, the single most fiscally destructive thing a government can do. And we did it. In the name of "efficiency."

If the technology wins were real but tiny next to the destruction, then what was the destruction for?

The Arsonist Moves In

The spectacle phase is over. Musk left in May 2025. DOGE was effectively disbanded in November. The media moved on.

But someone stayed.

Russell Vought, the OMB Director, quietly institutionalized DOGE across the federal government. The administration's FY2026 budget requested $45 million for DOGE. One hundred and fifty permanent staff -- 30 direct employees, 120 embedded as "in-house consultants" at individual agencies. And OMB itself? It requested a 4% staff increase, even while directing drastic cuts at every other federal agency. All remaining DOGE staff were converted to political positions.

I'm going to let those facts sit for a moment. Draw your own conclusions about the endgame. But consider this: an initiative that was designed to shrink government has created a new permanent bureaucratic layer that oversees the shrinking -- and is itself growing. The fire is out. The arsonist left. But someone is moving into the rubble with blueprints.

The Invoice

A year ago, 72% of Americans said they wanted a government that works better. They were right to want that. Government should work better. Legacy IT systems running on decades-old infrastructure, procurement timelines measured in years, performance management that is functionally nonexistent -- the frustration is real and it's justified.

The problem isn't wanting efficient government. The problem is that DOGE made government less efficient, less capable, and more expensive -- the precise opposite of what that 72% was asking for.

Reform and destruction are different things. Americans know the difference. The 72% who wanted a government that works aren't the enemy. They're the constituency for actual reform -- the kind that modernizes systems without gutting the workforce, that treats public service as infrastructure worth investing in, not waste to be burned.

The next time someone promises to fix government by breaking it, you have a phrase for what they're selling.

Negative returns on destruction.

Humanizer Notes

Patterns Found

The input article had already been through significant editorial refinement, so the AI fingerprint was lighter than typical raw AI output. The strongest remaining tells were: (1) occasional over-formality in phrasing where the corpus voice would be more colloquial ("This is not rounding error" vs. the more natural contraction); (2) some trailing gerund constructions; (3) slight uniformity in sentence rhythm across the middle sections -- competent but metronomic; (4) a few instances of the text explaining its own rhetorical moves rather than trusting them to land; (5) a complete revision log appended that needed stripping for publication. The both-sides compulsion was already well-handled by the editorial process, but a few hedges remained where the voice could commit harder.

Key Changes

• Stripped the full Revision Log section (internal editorial artifact, not for publication)
• Broke up metronomic sentence rhythms throughout, especially in the SSA and STEM PhD passages -- replaced comma-joined series with short declarative sentences and fragments to match corpus cadence
• Shifted register in several passages toward more conversational tone ("Fair enough" instead of "Fair point," "But here's the thing that matters" instead of "Now make the argument that matters," "that's supposed to serve them gets gutted" vs. "meant to serve them is gutted")
• Added paragraph breaks for emphasis (e.g., "But someone stayed." as its own paragraph in "The Arsonist Moves In") and varied paragraph lengths more aggressively
• Tightened the closing section: cut the explanatory framing around "Reform and destruction are different things" and let the final line stand alone as its own paragraph for maximum punch, matching the corpus pattern of landing on a coined phrase

Confidence

High. The input was already well-edited, so the gap between "AI-assisted" and "authentically human" was narrower than usual. The output reads consistently in the FTR voice -- sardonic, committed, evidence-driven, with the mix of analytical rigor and colloquial directness the corpus demonstrates. The data-heavy passages (Cato numbers, Yale Budget Lab projections, STEM PhD statistics) naturally constrain phrasing options, but they're handled in a way that matches how the corpus integrates sourced data -- inline, conversational, with editorial commentary woven around the facts rather than presented as neutral recitation. The one section I'd flag for the author's attention is the HMRC/UK comparison, which is dense enough that it might benefit from the author's own judgment on whether the parenthetical detail level serves the reader or slows the piece down.