Month: March 2024

From Jonathan Cohen, an expert on lotteries:

Lotteries are run by state agencies, and they’re exempt from truth in advertising laws.

Huh. This is why, for some reason, state lotteries are allowed to advertise jackpots as the sum of 30 years of payments, rather than the amount they'll actually give you tomorrow if you win.

Which is pretty strange if you think about it. “No one would let you state it in that form if it were a financial product,” says Charles Clotfelter. But we're all so used to it that we barely even notice it anymore.

Bob Somerby was watching TV the other night and heard something interesting:

We have this asylum law where, if you get to the United States, we are going to hear your asylum claim.... The president cannot fix that. And as long as that's the rule—that, get here and you can stay as long as you say the magic words: "I have a credible fear of returning to my country"—we're not going to fix the border.

Bob thought this was interesting because he hadn't ever heard anyone say this straightforwardly before. Certainly not President Biden.

Which.......might be true? I've written about asylum before, but usually I just assume everyone knows it's a problem and then move on from there. So in case you don't, here it is in plain words: If you get across the border and claim asylum, international law—long ago incorporated into US law as well—requires that we hear your case. You get to stay until we do.

So how big a problem is this? Here's a chart that tells the story:

Up through 2014 asylum was a minor part of border enforcement. But starting in 2015 it became a bigger and bigger share of our migrant numbers.¹ In 2018 nearly every border crosser from the previous year applied for asylum.² ³

At that point it leveled off while ordinary illegal immigration rose, and then in 2021 it began skyrocketing again. In 2023 about 800,000 people applied for asylum.

But— it's worth noting that even this huge number is only about a third of the total migrant population. Asylum has indeed become an ever growing problem, but ordinary illegal immigration is still a much bigger one.

¹A note about language: if you cross the border without permission, you're an illegal immigrant. But if you show up at a port of entry and apply for asylum you aren't. Applying for asylum is entirely legal. So what do we call all these people? I'm calling them migrants here for lack of a better word, since this generally implies anyone who's entered the country irregularly without applying for the appropriate visa.

²You have up to a year to apply for asylum once you're on US territory. For that reason, asylum claims usually correlate with the previous year's migrant numbers.

³In case you're curious, roughly half of all asylum claims end up in court and about half of those are approved. So, give or take, about a quarter to a third of all asylum seekers are successful.

Hilbert in his box.

I can hardly stand to pass this along, but here it is:

School-age children affected by the water crisis in Flint, Mich., nearly a decade ago suffered significant and lasting academic setbacks, according to a study released Wednesday, showing the disaster’s profound impact on a generation of children.

....The learning gap was especially prevalent among younger students in third through fifth grades and those of lower socioeconomic status. There was also an 8 percent increase in the number of students with special needs, especially among school-age boys.

....But researchers were puzzled to find that children who weren’t directly exposed to contaminated water at home still faced academic challenges, suggesting there were society-wide ripple effects.

This is from the Washington Post a couple of weeks ago. There's only one problem: it's completely wrong.

I don't blame the Post for this. The study they relied on really did say all this. The problem is that it left out one big thing: At the same time as the Flint water crisis, Michigan changed its educational testing. Scores went down because of the test change, which is why it affected everyone. For example, here's a chart from a paper published in 2023:

Scores dropped everywhere in Michigan when M-STEP was adopted. Nothing special happened in Flint.

So why the increase in special ed kids? What most likely happened was a nocebo effect. This is sort of the opposite of a placebo: it's when people experience something because they were told it might happen. That's the conclusion of the 2023 paper:

Between 2011 and 2019, including the 2014-15 crisis period, the incidence of elevated blood lead in Flint children (≥ 5µg/dL) was always at least 47% lower than in the control city of Detroit.... There is actually an inverse relationship between childhood blood lead and special education enrollment in Flint.

This study failed to confirm any positive association between actual childhood blood lead levels and special education enrollment in Flint. Negative psychological effects associated with media predictions of brain damage could have created a self-fulfilling prophecy via a nocebo effect. The findings demonstrate a need for improved media coverage of complex events like the Flint Water Crisis.

This just kills me every time it comes up. I hate hate hate having to sound like I'm downplaying the effect of lead exposure on kids, which is a serious problem. But the facts are stubborn: the Flint water crisis didn't last very long; blood lead levels in Flint kids changed only modestly; and there's very little reason to think it had any serious or widespread effect on IQ or school performance.

But years of yelling and screaming about how Flint kids have been ruined for life has probably convinced a lot of them that they're ruined for life. With rare exceptions, they aren't, and they never should have been told that.

The Census Bureau has released some new survey data about AI use and it's pretty interesting. Overall use of AI is still fairly low: about 5.4% of firms report using it now, expected to rise to 6.6% in autumn. Of those firms who report using AI, here's what they say they're using it for:

All told, this means that in the near future about a third of AI will be used to replace workers and another third will be used to replace existing computer systems.

Here are the top ten industries by expected use of AI in the near future:

Food service and retail are fascinating: both report fairly limited use of AI right now but expect huge increases over the next few months.

Finally, here is overall AI use by state:

Colorado is #1. I wouldn't have guessed that. Mississippi is #50. I would have guessed that.

Via the New York Times, here's an interesting look at the political/cultural orientation of various AI models that are publicly available and widely used. It comes from a paper by David Rozado, who did the testing.

On the left are the results from base, untrained AI engines made by OpenAI and Meta. They're pretty centrist.

On the right are the results from various commercial AIs that have been fully trained. These are the products you actually use. They all tend toward lefty libertarianism, but some more than others:

The model furthest to the left is Google's Gemini, which is no surprise. Anthropic's Claude is the closest to centrist of the best-known models. Both GPT-4 and Twitter's Grok are close, but slightly more lefty.

Rozado also showed that it's fairly easy to train an AI to be reliably liberal or conservative. You just have them read the appropriate books and magazines.

Aside from Google, the major chatbot models aren't especially ideological, but they're definitely not conservative. Is that because of their Silicon Valley heritage or just because our written culture leans a bit to the left? Or something else? My own guess is that it's partly both of those plus one other thing: all the models are explicitly trained to not be racist/sexist/hateful/etc. Conservatives may not want to hear this, but like it or not, that pushes them away from the modern right wing.

After last month's spike, PCE inflation came down in February:

Headline PCE was down only a bit, but core PCE—which the Fed focuses on—was down substantially to 3.2%. This is pretty good news since it suggests that last month's spike was probably pretty transitory.

On a more conventional year-over-year basis, headline PCE was 2.5% and core PCE was 2.8%.

NOTE: Oddly, inflation in goods was up 6.1% on an annualized basis, a big spike from last month. This was largely because of gasoline prices. Conversely inflation in services was up only 3.1%, a big decrease since last month. Go figure.

Here's an odd one. A few weeks ago the CFPB issued a new rule substantially cutting late fees on credit card payments. The US Chamber of Commerce quickly sued. They unsurprisingly chose to file in the Fort Worth division of the Northern District of Texas, home to judges Reed O'Connor and Mark Pittman, who are both well known for regularly ruling in favor of Republicans.

But a funny thing happened: Pittman got the case and has determined that suing in Texas was ridiculous. Rather, the case belongs in Washington DC since that's where the CFPB, the Chamber of Commerce, and most of the lawyers reside:

Venue is not a continental breakfast; you cannot pick and choose on a Plaintiffs’ whim where and how a lawsuit is filed.... The Rule at issue in this case was promulgated in Washington D.C., by government agencies stationed in Washington D.C., and by employees who work in Washington D.C. Most of the Plaintiffs in this case are also based in Washington D.C. and eighty percent of the attorneys in this matter work in Washington D.C.

....As far as this Court can discern, not one of the banks or credit card companies directly affected by the future Rule is located in the Fort Worth Division.... This case does not belong in the Northern District of Texas and certainly not in the Fort Worth Division....The Court concludes that this case should be and is hereby TRANSFERRED to the United States District Court for the District of Columbia.

Maybe folks are really getting the message that judge shopping is now a thing of the past. Very cautiously, I'd count this as good news.

Via David Dayen.

Just in case you're curious, here's the annual rate of bridge failures in the US:

There's no trend either up or down. It's perhaps also worth noting that virtually none of these failures were due to age or bad design. The vast majority failed due to floods, being hit by trucks, being hit by barges, or by construction work.

The LA Times reports today on the impact of raising the fast-food minimum wage in California to $20 per hour:

Chipotle, McDonald’s, Starbucks, Jack in the Box and Shake Shack are planning to raise menu prices. Fast-food franchisees are laying off employees or cutting their hours. Smaller independent business owners, meanwhile, worry their workers will bolt unless they also increase wages.

Think what you will about burger flippers earning $20 an hour, but the median wage of fast-food workers in California is currently about $18 an hour.¹ So this is an average increase of 12%. And the overall impact? "David Neumark, a minimum-wage expert at UC Irvine, estimated that overall prices will rise between 2.5% and 3.75%."

A 12% raise is substantial, but the market would have produced the same thing in three or four years. This is honestly not that earth shattering.

¹The May 2023 figure from the BLS was $17.63. Fast food wages since then have risen 2.5%, bringing it to $18.07 per hour.