Everyone in DC is highly skeptical of DOGE. They will point out that DOGE itself is an advisory committee with no actual legal power. They will point out that most of the spending that comes out of Washington is not made up of $25,000 studies on the sex lives of racoons and $10,000 grants to LGBT organizations in Honduras, but of enormously popular programs like Social Security and Medicare, that most of these programs just send checks which is intrinsically efficient, and that many programs that could be described as wasteful also have powerful constituencies to defend them, which makes it unlikely that an unelected advisory body with no formal legal or administrative powers could dent it.

But…

What if that's wrong?

What if, ackshually, a huge chunk of Federal government spending really is a bunch of fraud, waste, and abuse? And what if, by using clever tools like AI, you really could cut massive chunks of Federal government spending? Without significantly impairing programs that are needed or that are popular or that have important constituencies protecting them? Free money!

It actually makes sense, if you think about it. Let's be honest. If you live in Washington, you know a lot of parts of the Federal government are run incompetently. And you know some are run corruptly. (Sorry, "corruptly.") We should expect significant bits of waste, fraud, and abuse.

Well, according to a GAO report last year, the Federal government loses $233 billion to $521 billion annually to fraud. That's $2 to $5 trillion over ten years (your correspondent is good at maths). That's a big number, even for the US Federal government. Improper payments (a different category from fraud) totaled $161 billion last year.

We were fascinated by this interview by the good people at the Foundation for American Innovation, with Linda Miller, co-founder of the Program Integrity Alliance, an organization that deals with these issues.

So we began to read up on this. Basically, it seems, government fraud and waste stem primarily from systemic inefficiencies in how agencies share and verify information. Fragmented data systems, outdated privacy regulations, and bureaucratic silos create an environment where agencies cannot effectively coordinate or cross-reference information. This structural problem became particularly evident during the COVID-19 pandemic, when programs like the Paycheck Protection Program relied heavily on self-certification without adequate verification capabilities, leading to an estimated $200 billion in potentially fraudulent disbursements from the Small Business Administration alone.

A second major driver is the persistent underinvestment in modern fraud prevention infrastructure and capabilities. Many government agencies still operate with outdated IT systems and lack access to sophisticated data analytics tools that are standard in the private sector. This technological gap, combined with insufficient staffing and funding for oversight functions, leaves agencies ill-equipped to detect and prevent increasingly sophisticated fraud schemes. The problem is compounded by limited access to critical data sources that could help verify payment accuracy, with over 66% of improper payments in FY 2023 attributed to agencies' inability to access necessary verification data.

Perhaps most fundamentally, government agencies face a cultural and incentive structure that often prioritizes speed of disbursement over payment accuracy and program integrity. Performance metrics typically focus on operational efficiency rather than fraud prevention, while anti-fraud measures are frequently viewed as obstacles to an agency's core mission rather than essential components of good governance. This mindset, coupled with insufficient accountability at leadership levels, creates an environment where fraud prevention becomes a secondary consideration rather than a strategic priority. The result is a compliance-oriented approach that fails to proactively address systemic vulnerabilities in government spending programs.

According to Miller, modern data analytics tools, such as machine learning-based anomaly detection and predictive modeling, can transform how government agencies prevent fraud and improper payments. Specifically, agencies could implement real-time screening systems that automatically verify payment eligibility across multiple databases before funds are disbursed. For example, an expanded version of Treasury's "Do Not Pay" platform could cross-reference Social Security numbers against death records, check business tax ID numbers against corporate registries, verify addresses against postal databases, and validate bank account ownership - all within seconds. This type of automated verification, already standard in private sector banking, would have prevented much of the estimated $200 billion in fraudulent COVID relief payments that relied solely on self-certification.

Beyond basic verification, advanced analytics platforms can detect subtle patterns that indicate potential fraud. Natural language processing could scan loan applications for suspicious similarities suggesting organized fraud rings. Machine learning models could analyze historical payment data to identify unusual claim patterns or statistically improbable concentrations of benefits. Network analysis tools could map relationships between entities to uncover hidden connections suggesting collusion or identity theft. These capabilities could be centralized in an expanded version of the Pandemic Analytics Center of Excellence (PACE), giving all agencies access to sophisticated fraud detection while eliminating the need for each to build their own systems. However, successful implementation requires both specialized data scientists who can develop and refine these tools and updated data-sharing agreements that enable agencies to pool information while maintaining privacy protections. The Social Security Administration's continuing disability reviews demonstrate the potential return on such investments, saving $9 for every dollar spent on enhanced verification processes.

This sounds like exactly the sort of thing DOGE can, and should do, and we should all be very grateful if it does.

Policy News You Need To Know

#TheEconomyStupid — The new monthly jobs report came in well above expectations, with 256,000 new jobs created in the last quarter, against a forecast of 165,000, and a 4.1% unemployment rate versus a forecast of 4.2%. President Trump will take office with the wind at his back.

#VibeShift — BlackRock left a major climate group, as part of a broader trend of Wall Street firms exiting so-called green investing.

#PublicHealth — Former FDA commissioners Scott Gottlieb and Mark B. McClellan are in the medical news site news STAT News with an important article in favor of disease surveillance tools.

#TheScience — Don't snicker: Paper argues that the productivity of researchers has fallen in large part because "the average ability of researchers has fallen substantially" due to dumber people selecting into research careers. Welllllll…

#Immigration — Remember the Christmas blowout over H-1Bs? The Center for Immigration Studies' George Fishman has some very sensible, practical, ideas for H-1B reform. The two main ones: institute a wage floor that is set at the average wage for each occupation in each area of employment; allocate visas first to those individuals promised the highest salaries. These are very good ways to address the two main things that restrictionists have against H-1B as it currently exists, namely bringing down current American workers' salaries and bringing in workers who are supposedly "high-skill" but not actually that high-skill.

#Immigration — Speaking of the good folks at CIS, today, we learned that, both hilariously and sadly, CIS runs a "National Security Vetting Failures Database." Yes, it's real. Anyway, we learned this because an Egyptian student was arrested for plotting to bomb the Israeli consulate in New York, and adjudicators missed evidence of online extremism. Oh well. Who cares about Islamic terrorism in New York, right?

#Tech — A case on the constitutionality of online age verification laws is pending before the Supreme Court. At Law & Liberty, here's an interesting symposium on that question.

#LGBT — Yet another court defeat for the Biden Administration's pro-trans Title IX rewrite.

Friday Essays

You may have heard about the "loneliness epidemic." This essay by Derek Thompson in The Atlantic is a deep dive into this phenomenon, including both anecdote and data.

Fascinating reporting from Pirate Wires on how Soros-backed operatives took over key roles at Wikipedia.

"Mass. Exodus" — We love a good pun-based headline, and this essay in The New Atlantis is rewarding beyond its headline: "Massachusetts is one of the richest states in the country — because it’s pricing out its own middle class. Why did the state stop building enough to house them?"

This was a fascinating essay by C. Jarrett Dieterle in the current issue of National Affairs on the future of the American restaurant.

Interesting article by Justin Vassallo in Compact on how the arrival of Silicon Valley bros in the Trump coalition represents "The Revenge of the Neoliberals"

The ever-subtle and provocative Adrian Vermeule has some recommendations for a "Common Good Presidential Administration"