Kevin Drum

As you know, I've spent a long time arguing that the Fed's interest rate hikes aren't responsible for lower inflation. This is because inflation started to ebb within a few months of the hikes, and that's way too fast. It takes a year or two for interest rate changes to affect inflation.

Well, it's been a couple of years now, so where does that leave me? Still puzzled. You see, interest rates don't affect inflation magically. They slow down the economy by making loans more expensive, and a slower economy then brings down inflation. But that never happened. The economy never slowed down:

Average GDP growth since the interest rate hikes began has been about 3.0%. That's more than it was before the pandemic, which was a pretty strong growth era itself.

So what happened? This was well after stimulus spending had mostly dried up and wasn't affecting anything. So how is it that a large and sharp increase in interest rates didn't slow down the economy? Does anyone know?

I have a potentially stupid question to ask about this. I mentioned a few months ago that the Fed no longer controls interest rates via open market operations—that is, by buying and selling treasury bonds on the open market. Instead, it's mostly switched to the much easier method of changing the rate it pays banks for the reserves they keep at the Fed. If the Fed is paying 5.25%, no bank will loan out money for less, so interest rates automatically go up.

But is it possible that high interest rates have a different effect depending on how they're created? In other words, maybe higher interest rates without any open market operations behind them have an attenuated effect on the economy.

That sounds kind of stupid, but there's got to be something going on. More prosaically, maybe the fed hikes were just too small to have much of an effect. It was only a year ago that real interest rates went above zero, and even now real rates are only about 2%. In the past it's taken real rates of around 4% to touch off a recession.

Either way, it's hard to believe that no one really knows what's going on. How is it that the Fed hiked rates and produced not a soft landing, but no landing at all?

Test your knowledge:

Q: How long does it take the FDA to approve new drugs?

A: About seven or eight months on average.

No, no, wait. It takes years. Everyone says it takes years. What nonsense is this?

It's not nonsense. Once a New Drug Application is submitted, the FDA typically turns it around in a few months. What's more, studies have repeatedly shown that the FDA is generally faster than European, Canadian, and most other regulatory agencies around the world. They also approve more drugs than the other agencies.

What does take a long time is clinical trials. The entire process of discovery and testing of new molecules takes anywhere from 10-15 years:

So if you want to speed up new drug approvals, the place to focus is not the relatively short final stage but the tortuous clinical trial stage. Here's an estimate of how many drugs make it through the various development steps:

Lots of people complain about the FDA being too slow, and in some cases there's something to this.¹ But the FDA is careful for a reason: although you may not hear about it a lot, most drugs are failures. As the chart above shows, fewer than 10% of all candidate drugs end up making it through the entire gauntlet. The rest either don't work or turn out to be unsafe. The only reason we know this is because the testing regimen is so strict.

But there's one stage where almost everything makes it through: Phase 1 trials. This prompts me to wonder why we bother with it. Why not combine Phase 1 and Phase 2 and cut a couple of years out of the whole cycle?

I'm sure there's a good reason not to do this, but I can't figure out what it might be. I suppose that's because the answer is so obvious that no one ever bothers asking the question. But I'm asking. Why bother with a step that almost never uncovers serious problems?

¹This is probably more true for medical devices than drugs, but there are certainly cases of drugs that have taken longer to approve than they should have.

UPDATE: The answer appears to be pretty simple. Phase 1 trials are done on healthy volunteers to check for safety. Only if healthy patients can tolerate the drug is it given to sick people in the Phase 2 trial.

The House has finally passed three separate foreign aid bills. Here's how the money breaks down:

• Ukraine: $61 billion
• Israel: $17 billion
• Gaza humanitarian aid: $9 billion
• Taiwan: $8 billion

In addition, the House passed a bill that sanctions Iran and Russia and would force the sale of TikTok. All of the bills go to the Senate next.

The Ukraine vote was 311-112, and that was the closest of the four.¹ All of this was massively popular and bipartisan, but it still took months just to get a vote.

And yet—the bills did eventually pass. It's appalling that it took so long thanks to a small band of malcontents, but they did pass. Somehow, in our usual chaotic, backhanded, slapdash way, the United States once again has managed to do something big. This keeps happening despite everything. We keep saying that the country is ungovernable these days, but in the end, usually after thrashing around for an embarrassingly long time, we govern.

So what's the point? As best as I can tell, the end result is this: (a) we get as much done as we ever have, but (b) a lot more people get pissed off about it. I don't even know if this is a conscious strategy, but it's what the Republican Party has been all about ever since Newt Gingrich took over.

¹Although it's worth noting that Republicans voted against it, 112-101.

Every year an outfit called Climate Action Tracker publishes an estimate of how much warming we're on course for by the year 2100. Cipher, a news site focused on climate change, has collected those estimates for the past decade and made a chart out of them:

The projected amount of warming has dropped from 3.7°C to 2.7°C. That's surprisingly good news. I've had a pretty definitive opinion that we were making essentially no progress, but it looks like I was wrong. The Paris Agreement has had a real impact.

There are, of course, some caveats. First, there's no guarantee that everyone is going to do what they've promised. Second, we've plateaued at 2.7°C for the past few years, and that's still an alarmingly high number.

Still, progress is progress. The more we make, the less pressure there will eventually be for geoengineering solutions. And if we do end up shooting sulfur aerosols into the atmosphere, it means we can shoot less of it. More like this, please.

I just saw a tweet about Groq, an AI chatbot that uses Meta's LLaMA-3 engine. It's incredibly fast! You should try it!

So I did. And it is incredibly fast. But there are still problems. I asked it about myself and got five paragraphs in return. Four were basically fine, but there was also this:

Drum has a background in science and has a Ph.D. in physical chemistry from the University of California, Berkeley. He has worked as a researcher and writer for various organizations, including the University of California, Berkeley, and the National Science Foundation.

I wonder where this stuff comes from? Physical chemistry, of all things. I had a TA once in a chem lab who was a physical chemistry grad student, and it was a pain in the ass. The organic chem guys were pretty easygoing, but the physical chemists all demanded military precision. If I wanted that I would have joined the Army.

Anyway, speed is all very well, but these LLMs really need to work on their accuracy. Until it gets better no one is ever going to be able to rely on them.

POSTSCRIPT: It also said, "Kevin Drum is a respected and influential voice in the online media landscape." That's nice. Maybe I forgive it.

Over at Vox, Kelsey Piper lays down some stats on airline problems:

Are more planes having incidents than ever before? Or are we just hearing about more incidents? It’s mostly the latter.... The National Transportation Safety Board (NTSB), which investigates aviation incidents and accidents, lists 12 incidents on commercial aircraft in the United States so far this year. Last year, during the same time period, there were 13 such incidents.

So the only reason we're hearing so much about airline incidents is because journalists are hopping onto even the most minor mishaps. Should they?

Here’s the question I struggle with as a journalist: Do we have some responsibility not to write such stories?

Journalists take accuracy very seriously. Every journalist I know works very hard not to publish a story that’s wrong — and if they did, they’d feel obliged to issue a correction. But it’s much less clear what our obligations are with stories that are completely true, and about a subject readers want to read about, but that paint for those readers a misleading picture of the world.

I'm glad Piper is asking this question, but I wish she didn't struggle with it. I think the answer is clear: Either stop reporting every incident or make sure to provide context when you do. Every time. To me this is nothing tricky. It's just Journalism 101.

Unfortunately, this still leaves us with a question. Piper says there have been 12 incidents this year compared to 13 last year over the same period. I wasn't quite able to replicate her numbers, but since it's Boeing jets that are making news, I figured it was better just to search for Boeing incidents anyway.

The overall trend is down, but since the pandemic drop in 2020 there's been a steady rise. The number of incidents is now near the peak of the past 20 years. So maybe there's something going on after all?

In fairness, this is the problem with context. There are lots of ways to look at data and it's not always easy to figure out which is really the most descriptive. In this case, there's also the fact that "incident" means anything from an airplane exploding in flight to a flight attendant twisting an ankle in bad weather. It would take some real digging to figure out if Boeing jets are having more mechanical issues than past years, and it might even be impossible. After all, even if there's a mechanical failure, is it because the plane is defective or because a maintenance worker didn't do their job right? Who knows?

It's never easy, is it? Still, I think I'd pretty much agree with Piper: there's not a lot of evidence that Boeing is having any more flight issues than in the past, and that's true if you look at every incident or only serious ones.

According to the latest Harvard Youth Poll, liberals outnumber conservatives among college educated young people by 3:1. The precise split is 58%-21%.

In other words, among those who profess any ideology at all, only a quarter are conservative.

Do we have a housing shortage? How is that possible if the number of housing units per capita is at an all-time high—which it is?

Kevin Erdmann has a long, complicated post explaining that this is all about families having fewer children, but he never takes this to its logical conclusion. To recap: we don't really care about housing per capita—i.e., housing per individual person. More kids, for example, doesn't mean we need more housing. Instead we want to look at housing units per household. Here it is:

The number of housing units per household is higher than it was in 2001. It still doesn't look like there's a housing shortage.

But wait. This is circular reasoning. If housing is in short supply, it depresses household formation (more kids remain living at home, etc.). But the depressed number of households will then make it look like the housing-per-household ratio is still high.

That sounds complicated, and it's about to get worse. As it happens, the number of adults per household has, in fact, been going up. Is this by choice, or because it's been forced on people by housing shortages? One way to get a handle on this is to look at the housing-per-household ratio, but use the 2001 figure for household size throughout. Here's that:

Even using this measure, the ratio of housing to households is the same as it was in 2001. In other words, even if household size had stayed the same, housing would be about as abundant now as it was two decades ago.

Here's yet another way of looking at this:

Forget households and forget kids. Just look at raw housing per adult. As you can see, it's precisely the same today as it was in 2001.

It was surprisingly hard to get the data for this chart thanks to the Census Bureau's remarkably crappy collection of data about population—the thing that's supposed to be its prime purpose in life. I'd like to do this same chart for California, but it fills me with fatigue just to think about trying to wrestle out of the Census Bureau the adult population of California for the past 20 years. I'm sure it can be done, but God knows how.

If I had this data, I'm pretty sure it would tell us that California has a housing shortage but the rest of the country doesn't. That's what I usually seem to find. For now, aside from California, I continue to believe that the US doesn't really have a housing shortage. Maybe in a few hot cities here and there, but that's about it.

The jury has been chosen for Donald Trump's hush money trial and the judge has moved on to what's called a Sandoval hearing. This is where the prosecution explains the kinds of questions it would like to ask Trump if he decides to take the stand. For my own amusement and yours, here's a selection of excerpts from the New York Times coverage:

We are now hearing from Matthew Colangelo, a prosecutor with the district attorney’s office, who says that the civil fraud trial finding from the judge — “repeated fraud and illegality” — would be highly relevant in calling Trump's testimony into question.

....Colangelo is bringing up a moment during the civil fraud trial when the judge found that Trump had violated a gag order in that case. The judge in that case summoned Trump to the stand and found that he had lied, and said that his testimony rang “hollow and untrue.”

....Justice Merchan just read another judge’s determination aloud, from a case in which Trump sued HIllary Clinton. That judge found that the lawsuit was "completely frivolous, both factually and legally,” and “was brought in bad faith for an improper purpose.”

Justice Merchan reads another quote from the judge in that suit, in which he labeled Trump a “sophisticated litigant who is repeatedly using the courts to seek revenge on political adversaries." The judge said: “He is the mastermind of strategic abuse of the judicial process, and he cannot be seen as a litigant blindly following the advice of a lawyer.” Justice Merchan sounds inclined to allow prosecutors to cross-examine Trump on this matter.

....We are down to the last issue on the list, Trump’s 2018 agreement to dissolve his charitable foundation after he was sued by the New York attorney general’s office.

In other words, Trump is a congenital liar and nothing he says on the stand is worth the spit he uses to utter it. That sounds like a very reasonable thing to confront Trump with. It also suggests Trump would be insane to testify on his own behalf.

Here is Charlie using his IEEE-standard petabit-per-second NoseLink to gather information about his brother. Apparently everything checked out.