Should Medicare Cover Anti-Obesity Drugs?

It seems like we finally have anti-obesity drugs that are effective and come without deal-breaking side effects: GLP-1 inhibitors like semaglutide (Wegovy). But they are currently priced over $10,000 per year for Americans. Should insurance cover them?

So far Medicare has decided to cover these drugs only to the extent that they treat diseases like diabetes (which these drugs were originally developed to treat) and heart disease (Wegovy reduces adverse cardiac events by 20% in overweight patients with heart disease). Just based on the diabetes coverage, Medicare was already spending $5 billion per year on these drugs in 2022, making semaglutide the 6th most expensive drug for Medicare with prescriptions still growing rapidly. The addition of other indications for specific diseases, like heart disease coverage added last month, is sure to expand this dramatically, especially if trials confirm other benefits.

But with almost 3/4 of Americans now officially overweight, weight loss makes for a bigger potential market than any specific disease. Medicare currently spends about 15k per beneficiary for all medical care; if they actually paid for an 11k/yr drug for 3/4 of their beneficiaries, their spending could rise to 23k per beneficiary per year. The effect on Medicare Part D, which covers prescription drugs and currently spends about 2.5k per beneficiary per year, would be even more dramatic, with spending quadrupling. This would blow a huge hole in the federal budget, where health insurance already accounts for about 1/4 of all spending (and Medicare 1/2 of that 1/4).

Of course, the reality would not be nearly that bad. Not all overweight people would want to take a weight loss drug, even if it were covered by insurance; the side effects are real. To the extent people do take the drugs, the reduction in obesity could lead to lower spending on treatments for things like heart attacks. Rebates can already reduce the cost of these drugs to be less than half of their list price, and Medicare may be able to negotiate even lower prices starting in 2027. Key patents will expire by 2033, after which generic competition should dramatically lower prices. Competition from other brand-name GLP-1 drugs could lower prices much sooner.

Patents always come with a tradeoff: they encourage innovation in the future, but mean high prices and under-use of patented goods today. The government does have one option for how to lower the marginal price of a drug without discouraging future innovation: just buy out the patent. This would likely cost hundreds of billions of dollars up front, but this could be recouped over time through lower spending, while bringing large health benefits because the drug would be much more widely used if it were sold at a price near its marginal cost of production.

Of course, for now supply of these medications is the bigger problem than the cost. Even with the current high prices and insurers tending not to cover drugs of weight loss alone, demand exceeds supply and shortages abound. The manufacturers are trying to ramp up production quickly to meet the large and growing demand, but this takes time. Insurers like Medicare covering weight loss drugs wouldn’t actually mean more people get the drugs in the short run, it would simply change who gets to use them.

But once production ramps up, I do expect that it will make sense for Medicare to cover weight loss drugs. The health benefits appear to be so large that the drugs are cost effective even at current prices, and prices are likely to fall substantially over time. The big restriction I suspect will still make sense is to require that patients be obese, rather than merely overweight, since being “merely” overweight (BMI 25-29) probably isn’t that bad for you:

Source

Disclosure: Long NVO

Update 4/18/24: I started thinking about this question because of an interview request from Janet Nguyen at Marketplace. She has now published an excellent article on the subject that also includes quotes from John Cawley of Cornell, who knows a lot more than I do on the subject.

Social Media, Mental Health, and Young People

What’s the connection between social media use and mental health, especially among young people? You’ve probably heard a lot about this recently, in the media, by politicians, and among friends chatting about their kids. Lots of assertions are made, but there is also a bit of research on this topic. As someone who frequently uses social media myself, as well as a parent of young children, and a teacher that works every week with young college students, I am particularly interested in this topic.

Jonathan Haidt and various co-authors have been trying to catalog all the research on the topic and figure out if there is a connection between the decline in teenage mental health and the rise of social media use. Haidt also has a new book on this topic, as well as the decline of “free play” among kids, which I have not yet read but I’ve looked through his documents that contain all of the underlying and summaries of the research he is citing. I’ll read the book soon, as I’m certainly part of the intended audience (see the last sentence of the above paragraph). And while this research is very much outside of my area of expertise, my training as an economist has taught me how to read academic papers and to be convinced by evidence, so once again I’m very much the intended audience on this score as well.

Please read this post as my attempt to understand the evidence and start to form conclusions and/or critique what Haidt is saying. It’s a work in progress, and I’ll write more as I read and think more about it.

Continue reading

It’s Not Too Late to Get Your Eclipse Glasses for the April 8 Solar Eclipse

Surely you have heard by now that a solar eclipse is coming. As the April 8 date approaches, the media/social media coverage will likely rise to a roar. I think we all know that the experience of being in the path of a total solar eclipse is eerie and memorable – – birds and insects can fall silent as night-like darkness falls, and a noticeable chill may be felt in the air.

Maps abound of the eclipse path across North America. For the U.S.,  it starts in Texas around 1:30 Central time, traverses southern Indiana and northern Ohio around 3:10 Eastern and ends in northern Maine about 3:30. Here is a snip I took from this NASA map, where I zoomed in the Midwest/Northeast section, and traced in red the lines of 90% totality:

If you really want the 100% experience, and if you want it to last the full four minutes, you must be in a relatively narrow strip. And if you want to have good chance of not having clouds obscure the fun, you may need to fly to central Texas. Buffalo, New York is in the middle of the eclipse path, but it is a notoriously overcast place.

But lots of folks, including residents of Chicago, Toronto, and the Boston-Washington corridor, live within the zone of (nearly) 90% totality, where you can see the moon sliding across most of the sun’s disk over the course of a few minutes, and experience significant darkening. The next solar eclipse to touch the U.S. will not be until 2044, and that will be barely visible from three less-populated states, Montana, North Dakota, and South Dakota.

So, I suggest you take the opportunity to enjoy this one to the max. This absolutely entails using special glasses with filters designed for safe viewing of the sun. Do not even think of looking at the sun without such glasses, and be alert lest children pick up the wrong cues and try to look at the sun.

The good news is that eclipse glasses are still available. I ordered some from Amazon a couple days ago that arrived two days later, and I saw them for sale in Lowe’s today. I got some extra to share with random friends and strangers. This can be a great chance to interact with neighbors and children.

The price per pair of glasses varies a lot, so do comparison shop.  I look for ones that say “CE and ISO Certified” like these. Be safe and have fun!

Notes on ChatGPT from Sama with Lex

This is a transcript of Lex Fridman Podcast #419 with Sam Altman 2. Sam Altman is (once again) the CEO of OpenAI and a leading figure in artificial intelligence. Two parts of the conversation stood out to me, and I don’t mean the gossip or the AGI predictions. The links in the transcript will take you to a YouTube video of the interview.

(00:53:22) You mentioned this collaboration. I’m not sure where the magic is, if it’s in here or if it’s in there or if it’s somewhere in between. I’m not sure. But one of the things that concerns me for knowledge task when I start with GPT is I’ll usually have to do fact checking after, like check that it didn’t come up with fake stuff. How do you figure that out that GPT can come up with fake stuff that sounds really convincing? So how do you ground it in truth?

Sam Altman(00:53:55) That’s obviously an area of intense interest for us. I think it’s going to get a lot better with upcoming versions, but we’ll have to continue to work on it and we’re not going to have it all solved this year.

Lex Fridman(00:54:07) Well the scary thing is, as it gets better, you’ll start not doing the fact checking more and more, right?

Sam Altman(00:54:15) I’m of two minds about that. I think people are much more sophisticated users of technology than we often give them credit for.

Lex Fridman(00:54:15) Sure.

Sam Altman(00:54:21) And people seem to really understand that GPT, any of these models hallucinate some of the time. And if it’s mission-critical, you got to check it.

Lex Fridman(00:54:27) Except journalists don’t seem to understand that. I’ve seen journalists half-assedly just using GPT-4. It’s-

Sam Altman(00:54:34) Of the long list of things I’d like to dunk on journalists for, this is not my top criticism of them.

As EWED readers know, I have a paper about ChatGPT hallucinations and a paper about ChatGPT fact-checking. Lex is concerned that fact-checking will stop if the quality of ChatGPT goes up, even though no one really expects the hallucination rate to go to zero. Sam takes the optimistic view that humans will use the tool well. I suppose that Altman generally holds the view that his creation is going to be used for good, on net. Or maybe he is just being a salesman who does not want to publicly dwell on the negative aspects of ChatGPT.

I also have written about the tech pipeline and what makes people shy away from computer programming.

Lex Fridman(01:29:53) That’s a weird feeling. Even with a programming, when you’re programming and you say something, or just the completion that GPT might do, it’s just such a good feeling when it got you, what you’re thinking about. And I look forward to getting you even better. On the programming front, looking out into the future, how much programming do you think humans will be doing 5, 10 years from now?

Sam Altman(01:30:19) I mean, a lot, but I think it’ll be in a very different shape. Maybe some people will program entirely in natural language.

Someday, the skills of a computer programmer might morph to be closer to the skills of a manager of humans, since LLMs were trained on human writing.

In my 2023 talk, I suggested that programming will get more fun because LLMs will do the tedious parts. I also suggest that parents should teach their kids to read instead of “code.”

The tedious coding tasks previously done by humans did “create jobs.” I am not worried about mass unemployment yet. We have so many problems to solve (see my growing to-do list for intelligence). There are big transitions coming up. Sama says GPT-5 will be a major step up. He claimed that one reason OpenAI keeps releasing intermediate models is to give humanity a heads up on what is coming down the line.

Recovering My Frozen Assets at BlockFi, Part1. How Sam Bankman-Fried’s Fraud Cost Me.

Back in 2021, interest rates had been so low for so long that that seemed to be the new normal. Yields on stable assets like money market funds were around 0.3% (essentially zero, and well below inflation), as I recall. As a yield addict, I scratched around for a way to earn higher interest, while sticking with an asset where (unlike bonds) the dollar value would stay fairly stable.

It was an era of crypto flourishing, and so I latched onto the notion of decentralized finance (DeFi) lending. I found what seemed to be a reputable, honest company called BlockFi, where I could buy stablecoin (constant dollar value) crypto assets which would sit on their platform. They would lend them out into the crypto world, and pay me something like 9 % interest. That was really, really good money back then, compared to 0.3%.

On this blog, I chronicled some of my steps in this journal. First, in signing up for BlockFi, I had to allow the intermediary company Plaid complete access to my bank account. Seriously, I had to give them my username and password, so they could log in as me, and not only be able to withdraw all my funds, but see all my banking transactions and history. That felt really violating, so I ended up setting up a small auxiliary bank account for Plaid to use and snoop to their heart’s content.

I did get up and running with BlockFi, and put in some funds and enjoyed the income, as I happily proclaimed (12/14/2021) on this blog, “ Earning Steady 9% Interest in My New Crypto Account “.

BlockFi assured me that they only loaned my assets out to “Trusted institutional counterparties” with a generous margin of collateral. What could possibly go wrong?

What went wrong is that BlockFi as a company got into some close relationship with Sam Bankman-Fried’s company, FTX.  Back in 2021-2022, twenty-something billionaire Sam Bankman-Fried (“SBF”) was the whiz kid, the visionary genius, the white knight savior of the crypto universe. In several cases, when some crypto enterprise was tottering, he would step in and invest funds to stabilize things. This reminded some of the role that J. P. Morgan had played in staving off the financial panics of 1893 and 1907. SBF was feted and lauded and quoted endlessly.

For reasons I never understood, BlockFi as a company was having a hard time turning a profit, so I think the plan was for FTX to acquire them. That process was partway along, when the great expose’ of SBF as a self-serving fraudster occurred at the end of 2022. He effectively gambled with his customers’ money. This would have made him even richer if his bets had paid off, but they went sour, which brought everything crashing down.

FTX quickly declared bankruptcy, which forced BlockFi to go BK as well. SBF was eventually locked up, but so were the funds I had put into BlockFi. The amount was not enough to threaten my lifestyle, but it was enough to be quite annoying.

Sam’s parents are both law professors at Stanford who are now resisting returning to FTX’s creditors the  $32 million (!!!) in assets (cash and real estate) that SBF had given them out of FTX’s operations. Some of that $32 million they are hoarding is mine, since BlockFi needs to recover its claims against FTX in order to make BlockFi clients whole. Sam’s mother has denounced the legal judgment against her son as “as “McCarthyite” and a “relentless pursuit of total destruction,” which is enabled by “a credulous public.” One wonders what little Sammy imbibed in the way of practical ethics in that household of idealistic Stanford law professors – the “effective altruism” that the Bankman-Fried family touts is perhaps a gratifying concept, until it actually costs you something you don’t want to part with. But I digress.

BlockFi Assets Begin to Thaw

I got emails from BlockFi every few months, assuring customers that they would do what they could to return our assets. Their bankruptcy proceedings kept things locked, but now they are starting to return some money. A judge ruled in early 2023 that assets held by users in their BlockFi “wallet” belonged to the users and could be withdrawn. However, assets in the interest-bearing account (which is where my stablecoin was) technically still belong to the bankrupt company’s estate, and were not necessarily available for withdrawal. But now, following another legal agreement,  BlockFi is returning funds from the interest accounts. The problem is that you will only get some fraction of what you put in. Some YouTube commenters have complained they only got 10-25% of their assets, and no one seems to know if they will ever get more. Ouch.

I got an email from BlockFi saying that I have assets to claim, but I need to set up an actual independent crypto wallet to receive them. BlockFi will only transfer the actual coin, not the dollar values. So, I am in the middle of this process. It’s one thing to open a wallet, where you can transfer crypto coins in and out. It is another to exchange or monetize your coin; for that you seem to need an exchange.

I have chosen to go with Coinbase. It is not the cheapest alternative, but it seems to be the most solid U.S. based crypto exchange. I have opened a Coinbase account now. As with BlockFi, I had to go through Plaid (ugh) for the connection to my bank account.

Next thing I need to do is to open a Coinbase wallet, and try to connect with BlockFi, and see what I get back. I will post later on what happens there.

Update: I got scammed in this process, see here. My bad for clicking on a link in an email, instead of going to the official website for the link…

What the Superintelligence can do for us

These days, when I blog-rant about my everyday life, I have increasingly ended on the thought “AGI fixes this.”

Yesterday, I mused whether AGI would be my personal chef? : Where Can You Still Buy a Great Dinner in the US?

Would AGI help me match my clothes that I no longer want to humans who can use them, to cut down on pollution?: Joy’s Fashion Globalization Article with Cato

Would AGI make no mistakes about weather-related school closure?: Intelligence for School Closing

Can AGI book summer camp for me?

As a millennial woman working through my 30’s, I increasingly see social media posts from my friends like this one:

One of the difficult things about infertility, for my friends going through it, is the uncertainty. Modern medicine seems legitimately short on information and predictive analytics for this issue. So… AGI to the rescue, someday?

All I’m writing about tonight is that I have created a growing to-do list, over roughly the past year, for the AGI. Would something smart enough to do all of the above be dangerous? I wouldn’t rule it out. As pure speculation, it feels safer to have an AI that is specifically devoted to being a personal chef but which strictly cannot do anything else beside manage food. An AI that could actually do all of those things… would be quite powerful.

Here’s me musing about the AGI rising up against us, written after watching the TV show Severance: Artificial Intelligence in the Basement of Lumon Industries

Videos for Teaching Inflation in 2024

I’m teaching principles of macro this semester. Making macroeconomics sound important to students is partly about explaining that recessions are painful and significant.

As Alex Tabarrok says, “The Great Depression is Over!”  Maybe Gen Z can appreciate the significance of the Great Depression, but it is history. Gen Z has heard of the Great Recession, but keep in mind that a student who is 20-y-o in 2024 was 4 in 2008. It’s a weird one, but there has been a recession more recently. The Covid Recession is what I like to link to, when possible, in class.

To teach the inflation chapter this week, I’m using video clips that I’ll put up here as resources for others.

To start off the inflation chapter and bring in a more global perspective, I show: “Zimbabwe’s inflation rate hits triple digits”  This 2-minute news clip was produced by Al Jazeera. They talk about lending and policy in addition to retail price increases.

After we have gone through some definitions, I show two clips of an economic forecast that was recorded in 2021. I don’t usually show such long clips in class, but I’m relying on dramatic irony to make it interesting. The students know the path that inflation took from 2020 to 2024, but Dr. Doti in the video does not. I stop the video occasionally to point out connections to our textbook.

Chapman University’s 2021 Economic Forecast Update was presented virtually on Wednesday, June 16, 2021.

Dr. Jim Doti predicts that an unprecedented increase in the money supply after Covid will lead to inflation. He’s not right about everything, but that’s what makes it so interesting. Right after showing students the quantity theory of money equation, I can show them someone trying to apply it from about minute 25 to about minute 35. (don’t start the video from minute 1)

Then, I go back to my lecture and introduce the Fisher effect. Next, we watch about minute 38 to minute 43 of the 2021 forecast because of the direct connection of inflation to interest rates. Partly this just helps illustrate how messy the real world is.

Also, I pull from one of Jeremy’s 2023 posts to illustrate the long run neutrality of money. “The Rate of Inflation is Falling, But Prices are Still Rising (And So are Wages)

Covid Death Structural Breaks

xtbreak (STATA)

I found a new time series and panel data tool that I want to share. What does it do? It’s called xtbreak and it finds what are known as ‘structural breaks’ in the data. What does that mean? It means that the determinants of a dependent variable matter differently at different periods of time. In statistics we’d say that the regression coefficients are different during different periods of time. To elaborate, I’ll walk through the same example that the authors of the command use.

You can download the time series data from here: https://github.com/JanDitzen/xtbreak/blob/main/data/US.dta

The data contains weekly US covid cases and deaths for 2020-2021. Here’s what it looks like:

So, what’s the data generating process? It stands to reason that the number of deaths is related to the number of cases one week prior. So, we can adopt the following model:

That seems reasonable. However, we suspect that δ is not the same across the entire sample period. Why not? Medical professionals learned how to better treat covid, and the public changed their behavior so that different types of people contracted covid. Further, once they contracted it, the public’s criteria for visiting the doctor changed. So, while the lagged number of cases is a reasonable determinant of deaths across the entire sample, we would expect it to predict a different number deaths at different times. In the model above, we are saying that δ changes over time and maybe at discrete points.

First, xtbreak allows us to test whether there are any structural breaks. Specifically, it can test whether there are S breaks rather than S-1 breaks. If the test statistic is greater than the critical statistics, then we can conclude that there are some number of breaks. Note that there being 5 breaks given that there are 4 depends on there also be at least 4 breaks. And since we can’t say that there are certainly 4 breaks rather than 3, it would be inappropriate to say that there are 4 or 5 breaks.

Great, so if there are three structural breaks, then when do they occur? xbtreak can answer that too (below). The three structural breaks are noted as the 20th  week of 2020, the 51st week of 2020, and the 11th week of 2021. Conveniently, there is also a confidence interval. Note that the confidence intervals for 2020w11 and 2021w11 breaks are nice and precise with a 1-week confidence interval. The 2nd break, however, has a big 30-week confidence interval (nearly 7 months). So, while we suspect that there is a 3rd  structural break, we don’t know as precisely where it is.

Regardless, if there are three structural breaks, then that means that there are four time periods with different relationships between lagged covid cases and covid deaths. We can create a scatter plot of the raw data and run a regression to see the different slopes. Below we can see the different slopes that describe the impact of lagged covid cases on deaths. Sensibly, covid cases resulted in more deaths earlier during the pandemic. As time passed, the proportion of cases which resulted in death declined (as seen in the falling slope of the dots). It’s no wonder that people were freaking out at the start of the pandemic.

What’s nice about this method for finding breaks is that it is statistically determined. Of course, it’s important to have a theoretical motivation for why any breaks would occur in the first place. This method is more rigorous than eye-balling the data and provides opportunities to hypothesis test the number of breaks and their location. If you read the documentation, then there are other tests, such as breaks in the constant, that are also possible.


See this ppt by the authors for more: https://www.stata.com/meeting/germany21/slides/Germany21_Ditzen.pdf

See this Stata Journal article for more still: https://repec.cal.bham.ac.uk/pdf/21-14.pdf

Does GPT-4 Know How High the Alps Are?

I’m getting ready to give some public local talks about AI. Last week I shared some pictures that I think might help people understand ChatGPT, specifically:

My first thought is that GPT-4 was giving incorrect estimates of the heights of these mountains because it does not actually “know” the correct elevations. But then a nagging question came to mind.

GPT has a “creativity parameter.” Sometimes, it intentionally does not select the top-rated next word in a sentence, for example, in order to avoid being stiff and boring. Could GPT-4 know the exact elevation of these mountains, and it is just intentionally being “creative,” in this case?

I do not want to stand up in front of the local Rotary Club and say something wrong. So, I went to a true expert, Lenny Bogdonoff, to ask for help. Here is his reply:

Not quite. It’s not that it knows or doesn’t know, but based on the prompt, it’s likely unable to parse the specific details and is outputting results respectively. There is a component of stochastic behavior based on what part of the model weights are activated.

One common practice to help avoid this and see what the model does grasp, is to ask it to think step by step, and explain its reasoning. When doing this, you can see the fault in logic.

All that being said, the vision model is actually faulty in being able to grasp the relative position of information, so this kind of task will be more likely to hallucinate.

There are better vision models, that aren’t OpenAI based. For example Qwen-VL-Max is very good, from the Chinese company Alibaba. Another is LLaVA which uses different baselines of open source language models to add vision capabilities

Depending on what you are needing vision for, models can be spiky in capability. Good at OCR but bad at relative positioning. Good at classifying a specific UI element, but bad at detecting plants, etc etc. 

Joy: So, I think I can tell the Rotary Club that GPT was “wrong” as opposed to “intentionally creative.” I think, as I originally concluded, you should not make ChatGPT the pilot of your airplane and go to sleep when approaching the Alps. ChatGPT should be used for what it is good at, such as writing the rough draft of a cover letter. (We have great “autopilot” software for flying planes, already, without involving large language models.)

Another expert, Gavin Leech, also weighed in with some helpful background information:

  • the creativity parameter is known as temperature. But you can actually radically change the output (intelligence, style, creativity) by using more complicated sampling schemes. The best analogy for changing the sampling scheme is that you’re giving it a psychiatric drug. Changing the prompt, conversely, is like CBT or one of those cute mindset interventions.
  • For each real-name model (e.g. “gpt-4-0613”), there’s 3 versions: the base model (which now no one except highly vetted researchers have access to), the instruction-tuned model, and the RLHF (or rather RLAIF) model. The base model is wildly creative, unhinged, but the RLHF one (which the linked researchers use) is heavily electroshocked into not intentionally making things up (as Lenny says).
  • It’s currently not usually possible to diagnose an error – the proverbial black box. My friends are working on this though
  • For more, note OpenAI admitting the “laziness” of their own models. the Turbo model line is intended to fix this.

Thank you, Lenny and Gavin, for donating your insights.