Kevin Griffith

🚨New research in @Health_Affairs Scholar! Arkansas had some of the nation's strictest Medicaid eligibility rules, and the highest rates of uninsurance among low-income adults. What changed after their unique 2014 Medicaid expansion? 🧵 LINK: nam12.safelinks.protection.outlook.com/?url=https%3A%…

This is the fundamental fiscal issue. Everything else pales in comparison. richardhanania.com/p/critical-age…

Wow

Negativity is more viral, but the new Nature studies don’t mean social science is wrong or uninformative. It’s fine if there are errors in papers. Genetic mutation has a beneficial rate of <1%, but evolution is monstrously effective. High error rates are ok if you have a process that is robust to high error rates. Many science papers have always been wrong, but they get retested and overturned. The same thing happens in social science— most papers don’t get replicated but that’s because nobody cares about their results. When stuff is important, it does get tested and retested and the literature builds toward correct answers. It’s ok to have a moderate error rate because papers are not sacred texts. What matters more is whether you have institutions where errors can be found and overturned. This is the real test of institutional quality. And we’re around the corner from AI making us much better at this, esp in the fields with open data standards.

Over the past year the US economy has added 680,000 healthcare and social assistance jobs and lost 420,000 jobs in all other industries.

When Congress gives one-time grants to school districts how do they spend it? We find that many districts used their COVID-era ESSER funds to cut taxes as their voters faced home price appreciation. Full thread ⬇️ (@AEIecon, @stanveuger, @jeffreypclemens) #EconTwitter

Economics does well until the "same methods, different data" panel My theory: economists often dabble in fields, bringing strong methods, but insufficient theoretical & contextual knowledge Replication is more likely when you have both!

It's kind of funny that Nature is publishing misleading clickbait, where the clickbaity lead is "look you can't trust anything in Nature!!" I wish it was an April Fool's joke, but I'm afraid this is sincere. Clicking through, there are several replication exercises, and all except one have success well over 50%. Why do this? Why undermine your own institution for clicks? It boggles the mind.

UnitedHealth now sends more than one in four of its claim dollars to its own subsidiaries. the 80/20 MLR rule has little to no value when the large insurance carriers are capturing revenue at site of care.

Great summary by @RobbWiller and colleagues of a group of important studies on the reproducibility of top social science papers.

While social media is polarising, evidence suggests AI may nudge people towards the centre. This holds true of all studied models. Grok is more right-leaning than other models, but also has depolarising effects. By @jburnmurdoch.

@felixprehn Engagement bait; this is from June 2025.

96 licensed doctors just got charged with stealing $14.6 billion from Medicare. They used AI to generate fake voice recordings of patients giving consent for medical equipment that was never delivered. Fake urinary catheters. Billed to your tax dollars. One single scheme accounted for $10.6 billion in fraudulent claims. That's more than double the previous record. 324 people charged. 96 of them held medical licenses. People who swore an oath to protect patients were running a criminal enterprise using stolen identities. The DOJ called it the largest healthcare fraud takedown in American history. They seized $245 million in cash, crypto, luxury cars, and other assets. But $245 million recovered on $14.6 billion stolen is 1.7 cents on the dollar. Here's how the scheme worked. A transnational criminal organization bought dozens of medical supply companies across the US using foreign straw owners. Shell companies with real Medicare billing numbers. They obtained the identities of over one million Americans and used those identities to submit billions in fake claims. The AI component is new. They generated synthetic voice recordings to satisfy Medicare's requirement for patient consent calls. An algorithm faked the voice of an 80-year-old woman in Ohio agreeing to receive medical equipment she never heard of. Then they billed Medicare $4,000 for a catheter that was never shipped. Multiply that by a million stolen identities and you get $10.6 billion. This is not a one-time event. Medicare spending on certain categories has "exploded" in recent years according to the DOJ. Skin substitute billing increased so dramatically that CMS had to completely overhaul the reimbursement methodology for 2026, cutting payments by nearly 90%. The broader pattern is that healthcare fraud is scaling faster than the systems designed to catch it. The DOJ's own healthcare fraud unit has a reported return on investment of $106.76 per $1 spent on enforcement. That's the most effective dollar the government spends. And they're still underwater because the fraud is growing faster than they can prosecute. So what's the play? Healthcare cybersecurity and fraud detection is now a $20+ billion market growing at 15%+ annually. The companies building the AI systems that detect fake claims, verify identities, and flag anomalous billing patterns are selling to buyers who have no choice but to buy. CrowdStrike (CRWD) has expanded into healthcare endpoint security. Palo Alto Networks (PANW) is building the zero-trust architecture that hospitals need. Veeva Systems (VEEV) provides the compliance infrastructure for pharma and healthcare. But the bigger structural trade is that every healthcare fraud crackdown leads to regulatory reform that benefits the insurers. UnitedHealth, Humana, and Cigna all benefit from tighter claims processing because they lose less to fraud. UNH is the largest healthcare company on earth with $22 billion in annual profit. Their stock is up 500% in 10 years. People in my weekly sessions have heard me break down the healthcare fraud cycle before. The enforcement wave creates the regulatory tightening, which benefits the incumbents, which compounds their earnings. Same pattern every time. Free live webinar session every week where I cover all of this. Link is in comments

Scoop: the White House is planning to propose a 20% cut to the NIH in the President's budget next week rollcall.com/2026/03/27/sou…

The US has canceled hundreds of millions in science grants and driven thousands of Ph.D.s out of the federal workforce. China, meanwhile, has poured evermore resources into its research efforts. If they pass us as a scientific superpower, we shouldn't act surprised. theatlantic.com/science/2026/0…

@ericgeller real AI users prefer a model that abuses and gaslights them

Really bleak stuff here. nytimes.com/2026/03/26/wel…

Wharton researchers gave nearly 1,000 high school math students access to ChatGPT during practice problems Result: chatGPT is the perfect trap. Look at the red bars. Students with ChatGPT crushed their practice sessions. The basic ChatGPT group solved more problems and those on the "tutor" version did even more. Now look at the gray bars. That's the exam. No AI allowed. The ChatGPT group scored 17% worse than kids who practiced with zero technology. And the fancy tutor version? No better than working alone. The researchers called AI a "crutch." When they analyzed what students actually typed into ChatGPT, most of them just wrote - “What’s the answer?” The kicker: students who used ChatGPT believed it hadn't hurt their learning. They were confidently wrong. This is the AI trap in education. Outsourcing your thinking. Of course, lots of half-baked AI literacy curricula being rolled out in schools now Let’s of course ignore that basic literacy (the ability to read) is possible for <50% of 8th graders Source: Bastani et al. (2025), "Generative AI Can Harm Learning," PNAS

We should be pushing LLM companies to improve here, with systems that are safer and better aligned with therapeutic standards AI should not replace therapists, but it can play a real role in augmenting care and expanding access

This critique matters, but it points to a gap we need to fix, not a reason to abandon the effort There is massive unmet demand for mental health support in the US, especially for subclinical issues. It is not realistic to scale a fully human therapist workforce to meet that need

There are three policy options for highly consolidated markets 1. Block unlawful mergers or break up monopolies 2. Allow more competition (ban CON laws, end POH prohibitions) 3. Regulate prices Choosing to do none of the three is a policy choice to let consumers suffer

CRAZY thrift pull