Author: Kevin Drum

The NAEP has released brand new results for its long-term reading and math tests. The long-term test is a special version of the NAEP that strives to stay the same from year to year so it produces scores that can be compared over time. Here's the main takeaway:

Despite the pandemic, reading scores didn't go down much, but the scores in mathematics were fairly disastrous. Math showed a decline of nine points in three years, sending scores back to their 1990 levels. Despite their small absolute drop, reading scores were back to their 1975 levels.

Score drops were significantly different for different racial and ethnic groups:

Using the rough rule of thumb that 10 points equals one grade level, white students lost about half a grade level while Black students lost 1½ grade levels. Native Americans lost an astonishing two grade levels.

Similar trends could be seen among different performing students. The highest performing students lost 7 points while the lowest performing students lost 15 points. Students in the middle lost 8 points.

So scores were down and the racial gap went up. It doesn't get much worse than that. There's considerably more reason to panic now than there was for last year's scores, which I predicted kids would make up in the next test. I was wrong.

I promise I'm not stalking Matt Yglesis these days, but he just happens to have written on some interesting topics lately. On Tuesday he ran an interview with Jennifer Doleac, an expert in how to reduce crime. The fascinating thing about this is that from 100,000 feet it seems like an intractable problem. Nothing works. But it turns out that lots of things work, and they're all things we're familiar with. We just have to do them. Unfortunately, as you've guessed, there are lots of people who don't want to bother. They figure it's cheaper and easier to just toss lots of criminals into prison and give them ever longer sentences. After all, if you're cooped up in a cell at San Quentin, you're not on the street robbing little old ladies.

I'll get back to all that in a bit. First, though, some context. As we all know, crime in the US has fallen precipitously since 1991. It is now about half the number it was in the peak year of 1991:

But the number of police officers has increased. So if crime is down but cops are up, that means each cop has a lot less crime to worry about:

The number of cops per crime has nearly tripled over the past 40 years and the number of crimes has about halved. What's more, the decline of lead poisoned teenagers has produced kids who are fundamentally less vicious than they used to be. As a result, policing is safer today than it's ever been before:

This is background, partly to show that many of the suggestions to follow are quite feasible. My list of things we can do is paraphrased from Matt's interview with Doleac, and I was surprised to see that I was familiar with all of them but one (#6). Let's go through them.

1. Lead has reduced crime a lot over the past 30 years—and we could reduce crime even more by getting rid of more lead. This costs money, but we know how to do it.
   

   Investing in removing lead from such sources as topsoil, paint in older homes, and drinking water...would likely yield large reductions in violent crime in the future. There would be other benefits too, such as increases in cognitive function and academic achievement, reductions in mortality, and increases in fertility.
   .
   Lead remediation is costly. But violent crime is costly, too. Stephen Billings and Kevin Schnepel consider the effects of lead remediation on house prices. They estimate that each $1 spent on lead remediation generates $2.60 in benefits. [But there are other benefits] such as a reduction in the social costs to potential crime victims and a reduction in the level of law enforcement that would be required if criminal behavior falls. For this reason, the benefits of lead remediation are likely much larger than what Billings and Schnepel estimate.

2. As you saw in the second chart, we have lots more sworn police officers per criminal incident than in the past. All by itself this should make policing more efficient. Throw in technology like Compstat and it makes added policing even more efficient.
   

   One of the classic ways to increase the probability of getting caught is to put more police on the streets. There are these experiments with hotspot policing where you put out a cop on this street corner, but not on another street corner. It turns out crime goes down on the street corner where the cop is standing.

3. Whatever you think of it, we have a lot more camera surveillance in public places than we used to. Especially when paired up with facial recognition, this is a powerful investigative tool that wasn't available even 10 years ago.
   

   Putting CCTV cameras in subways for instance is effective, putting cameras everywhere. It's like the London, UK model of lots of cameras everywhere

4. It is now routine to take DNA samples not just from convicts, but from anyone who's been merely arrested. This has built up a tremendous collection of DNA samples that can be compared to suspects in both violent and property crimes.
   

   [In Denmark's DNA program] we see recidivism fall by about 40%. So it's just a huge reduction in the likelihood that you commit another crime in the future.... Surveillance tech, putting cameras everywhere, is pretty invasive and so we might worry about that more. But if it's effective. We might be willing to make that adjustment, right?

5. It's common to go easy on probation violators because the penalties for violations are relatively severe and don't seem justified. A better solution is to mandate short but sure punishments (perhaps a day or weekend in jail) for things like alcohol or drug use. The "sure" part makes it clear to probationers that they will be punished. The "short" part ensures that they won't lose their jobs or otherwise have their lives ruined because of a minor violation.
   

   There are a bunch of innovations in the supervision space. Especially for people who have criminal behavior that is a function of their drug or alcohol use. You can kind of ramp up the testing for that and have more certainty of consequences, so you do need to do these drug tests daily or every other day or wear a blood alcohol content monitor or something, and then if you are drinking when you shouldn't be you immediately go to jail for a night and that works. You see a dramatic reduction in drug and alcohol use and a reduction in criminal behavior because people know that there will be consequences for breaking those rules.

6. It turns out that large employers are generally willing to hire non-violent ex-cons. But what they want is a large insurance policy—say, $1 million or so—to protect them from liability if an ex-con injures a customer or fellow employee on the job.
   

   [A research group] partnered with some researchers to run an experiment to see what would convince the employers to be open to hiring people with criminal records....And so they tried a whole bunch of different stuff like maybe wage subsidies right? So the employment platform will pay half the wage....The treatment that seemed most effective — all of these things were effective to some degree — but the most effective one was providing insurance up to a maximum of $1 million. So covering the liability risk.
   .
   And that's way higher than current programs that exist in the government that will cover $5,000. That covers you for things like property theft. It turns out employers don't care about property theft. That's not what they're worried about, they're worried about the assault on the customer or fellow employee and so the million-dollar coverage takes care of that and so that had a huge effect.
   

7. In the US we hand down massively long sentences compared to most other countries. We could cut sentences in half across the board and benefit practically everyone. First are the criminals, who would spend less time in prison. Second is the public budget for incarceration, which could be cut nearly in half. Third, a shorter sentence makes it less likely that a convict will spend enough time locked up to learn the ropes and become a career criminal.
   

   If they're thinking about what the expected cost of committing a crime is — including that punishment — then making the punishment worse and worse, longer and longer prison sentences, should deter people from committing a crime. And so we now have a lot of evidence that’s like, that doesn't work very well.
   .
   Other countries now routinely use electronic monitoring for instance instead of short prison sentences and they get big returns from that. Do we see recidivism drop dramatically in those countries?...And it turns out yeah, so it becomes an empirical question but the empirical answer in the countries that do it says it’s a total win. And so we should do it more in the U.S.

8. Cognitive Behavioral Therapy may sound like a little too much woo, but it works well if it's conducted by a good therapist. And that's the rub: there just aren't enough good CBT therapists to scale this up. I wonder if this is something that AI might help with. Not now, but maybe in a couple of years.
   

   There have been nice randomized trials in schools with higher-risk students and with kids who are actually in juvenile detention facilities and enrolling them in this CBT program reduced recidivism, reduced future arrests, reduced violent arrests, and increased school engagement all of these good things. And then the question becomes can we scale this? And I think efforts to scale it have not worked as well partly because you have to dig deeper into your applicant pool of potential leaders of these therapy programs. So the leaders of those programs aren't as good as the ones you had before.

So that's it. Some of these are aimed at preventing initial crimes while others are aimed at keeping probationers from breaking the law again after they're let out. As the list makes clear, these are all straightforward and technically feasible and we have pilot studies and multi-country comparisons for all of them. The problem is that some are expensive (#1, #5); some have big constituencies that oppose them (#7); some are complicated and a pain in the ass to set up (#5); some produce moral objections (#3); some are relatively unknown (#6); and others have problems scaling up (#8).

So it's no picnic. But it's far from impossible either. It would require us to get serious about solving the crime problem, but it wouldn't require any kind of technological or criminological breakthrough. When can we get started?

Jane Rosenzweig writes in the LA Times today about the perils of AI:

Soon after ChatGPT was released, an artificial intelligence researcher from one of the big tech companies told me that I shouldn’t worry about how the technology would affect how students learn to write. In two years, she assured me, only aspiring professional writers would enroll in writing classes; no one else would need to write anymore.

I've got good news and bad news. The good news is that two years is a laughably brief timeline. The researcher in this anecdote is generally right, but it will take a good deal longer than that.

The bad news is that it's not just writing. In a decade or two, when AI and robotics become really good, there will be little motivation for humans to learn anything. Why bother when AI is everywhere and can be better than you at literally everything. I can see elementary school surviving so that we humans know the basics, which are handy to have quickly at hand, but five grueling years to get a PhD when any research you might do can be had in a few hours or days just by asking your friendly local AI? I don't doubt that there are some people so obsessed they'll do it anyway—with an AI as their advisor—but it's going to be a pretty small number.

Someday. That's really the only question. It's going to happen, we just don't know precisely when.

Like all of you, I've been hearing endless reports that border encounters are way down since President Biden eliminated Title 42 enforcement last month. Today the Border Patrol released the official monthly count for May, and I was surprised to see . . . nothing:

Total encounters were down by only 6,500 people from April, or about 3%. Needless to say, this is only a single month of data, which can cut both ways. Reality might be substantially higher or lower when we get border results for the rest of the year.

But there's another way of looking at this. Title 42 expulsions, which can be used by border officers to turn away asylum seekers if they're deemed a health risk, are way down—and yet total encounters have hardly changed at all. Here's a chart that shows all encounters except Title 42 ever since the policy began in March 2020:

This produces a somewhat different picture. Border encounters excluding Title 42 are up nearly 50,000 people in May, a 40% increase. I'm truly not sure what to make of this. The elimination of Title 42 obviously eliminated Title 42 (the remaining small number of Title 42 expulsions in May are from the beginning of the month, before it was revoked). Duh. But non-Title-42 encounters were up considerably. So does this mean that, measured correctly, border encounters are up or down? I'm not deep enough in the weeds of illegal immigration to know.

As an aside, it's worth noting that the partisan conservative take on the border is almost certainly wrong. Border encounters skyrocketed in early 2021, and this is almost certainly due to the lingering effects of Donald Trump's policies before Joe Biden had a chance to do anything significant.

Since then, border encounters have been up and down, but today they're about the same as they were in the middle of 2021. It's fair to say that Biden hasn't done anything to reduce illegal immigration or asylum claims, but it's almost certainly not true to say that he's done anything much to increase them either.

Is social media bad for teens? So far, the evidence suggests only a single serious negative impact:

Depressed girls who start using social media 5-6 hours a day are likely to become even more depressed.

As a single bit of survey evidence, there's also the internal Instagram study showing that Instagram had a positive effect on almost everything but a negative effect on girls' body image. Aside from this, as Claire Cain Miller says in the New York Times today:

Reviews of the existing studies on social media use and adolescents’ mental health have found the bulk of them to be “weak,” “inconsistent,” “inconclusive,” “a bag of mixed findings” and “weighed down by a lack of quality” and “conflicting evidence.”

Despite this, I admit that my inclination is to believe social media has a net unhealthy influence on teens. This is because: so many platforms use toxic targeting algorithms that prioritize outrage; screen time in general is too high; spiraling FOMO is addictive and anxiety inducing; and quick written messages tend to be nastier and more virulent than in-person conversation.

But I have no evidence for this, and much of it is probably because I'm 64 years old and hardly ever use social media. People like me—and we're the ones who most often disparage social media—are probably the least likely ones to have an informed opinion.

This is a barnyard goose, probably an African goose if my perusal of "goose breeds" on Google is anything to go by. The last time I looked we had four of them, all hanging around near our little artificial lake.

Well, he's guilty of two misdemeanor charges of paying his taxes late. He did pay the taxes, he just didn't pay them on time.

Biden Jr. also avoided prosecution on a sketchy gun charge.¹ He was instead admitted to a diversion program, which is pretty common for first-time offenders. Together, this apparently ends the Biden investigation completely, including all the frantic claims that Biden took bribes, paid off his father, got a Ukrainian official fired, etc.² Politico has the details:

"It is my understanding that the five-year investigation into Hunter is resolved," Hunter Biden’s attorney Christopher Clark said in a statement....The plea agreement is intended to be a comprehensive resolution of Hunter Biden’s potential legal liability in all matters investigated by federal authorities, a person familiar with the negotiations said.

As usual, it's worth noting that the investigation of Hunter Biden was overseen by a Republican who was appointed by Donald Trump. David Weiss spent five years on it.

Also as usual, Weiss was approved on a voice vote in 2017 and no one seemed to have any problem with his appointment at the time.

But yet again, as usual, Republicans are now crying foul. Donald Trump: "Our system is BROKEN!" Rep. James Comer: "The Department of Justice’s charges against President Biden’s son Hunter reveal a two-tiered system of justice . . . We will not rest until the full extent of President Biden’s involvement in the family’s schemes are revealed." Speaker Kevin McCarthy: "If you are the president’s son, you get a sweetheart deal." Rep. Doug LaMalfa: "There needs to be a thorough investigation of . . . the Justice Department’s handling of the investigation." Rep. Elise Stefanik: "It’s never been more clear that we have a two-tiered system of justice . . . House Republicans will not rest until the full illegal corruption of the Biden Crime Family is exposed." Rep. Mary Miller: "The cash & diamonds he didn't pay taxes on came from CHINA, RUSSIA & UKRAINE. No FARA foreign agent charges? TWO SYSTEMS OF JUSTICE!" Gov. Ron DeSantis: "If Hunter was not connected to the elite DC class he would have been put in jail a long time ago."

The Republican keywords for the day are "two-tiered," "sweetheart deal," "slap on the wrist," and "elite DC class." More to come, I'm sure.

¹ATF was reluctant to even work the case from the start. Hunter was on drugs, tossed a gun away in a dumpster but did it badly, and agreed to rehab. Anyone else probably would have been given a warning and let go.

²If you want all the grimy details about this hysteria, just click here.

New business formation was up 0.4% in May compared to April. It was up a respectable 3.5% since last year.

But that's not super interesting. A longer term look is:

Note that this is not net business formation. It doesn't account for businesses that were shut down during the pandemic. Nonetheless, new businesses were obviously opening up even while others were closing, and they opened in huge numbers.

Even more interesting is that this continued through 2021 and then flattened out at the new higher number. Four years after the Great Bankruptcy, we're still opening new businesses at the rate of nearly two million per year more than we did pre-pandemic. That's what the trendline suggests, anyway.

Is this because net business formation has stayed about the same, but the new business formation number is high because existing businesses are still closing down at high rates? Or are business closures down to normal and new businesses really are booming? Employment levels suggest the former: The number of workers today is nearly the same as the number of workers just before the pandemic. If employment is about the same, then total business formation is probably about the same too.

But this means that existing businesses are still closing down at a high rate even though COVID-19 has largely lost its bite. Why?

UPDATE: A regular reader emailed today with a Fed report that summarizes BLS data on business closures. I had forgotten this existed. It only goes through the end of 2021, but if it's accurate it suggests that there was one quarter of high business closures (Q2 of 2020) and everything has been steady since then at the usual rate of about 1 million closures per year:

If this is right, it means just the opposite of what I concluded above. Closures have stayed about the same, which means that the boom in business formation is real.

New housing starts skyrocketed in May by nearly 300,000 units from April. That's surprisingly good, but it comes on the back of a steady decline over the past year. I sometimes struggle to pick out the best way to show chart data in the most informative possible way, and this was one of those times. Percentage growth is misleading because the April number was close to zero, while month-over-month numbers are hard to read (either in units or percentages) because the series is so noisy. Willy nilly, I ended up doing a rolling 6-month average of the monthly change in units:

The basic story is the same no matter what series you use. The chart is just a way of illuminating it. Housing starts were high during the 2021/22 housing boom and then plummeted during the 2022/23 crash. But in May they suddenly spiked way upward.

For a little bit of context, here's a comparison of single-family starts vs. apartment starts:

It's only one month, so the change in single-family starts might be a fluke. But the jump in May was its second-biggest during the entire past decade. This suggests newfound optimism about the housing market, which is hard to attribute to anything other than a sense among investors that interest rates will fall over the 12 months or so.

Don't worry, everything is fine.

If you've noticed that blogging has been a little sparse for the past few days, and dropped to nearly nothing today, it's not just because the news has been a little dull. It's because I feel crummy.

A week ago I got an upset stomach. The next day I developed a peculiar and intense pain in my upper back as well as severe headaches off and on. Plus my sleepiness and fatigue went off the charts.

Last night the upper back pain got so bad I couldn't sleep. I tried treating it with Advil, morphine, and dilaudid, the last of which is a very powerful opioid. None had more than a slight effect. The stomach I tried treating with Pepcid, Zofran, and Tums. No dice.

This morning I started breathing heavily, so I finally went into the ER to get checked out. Luckily, we were pretty thorough in making sure the ER doc knew that I had just gone through a lengthy CAR-T procedure, and a CT scan showed lytic lesions in a couple of my vertebrae. The lesions are caused by multiple myeloma and have been there for nine years now, but even so it was enough to get the doctor to contact my transplant doctor from City of Hope, and he had the answer: inflammation. Unfortunately, he also had the solution: dexamethasone. So now I'm back on the Evil Dex™ for a couple of weeks. Which also means I'm back on the Ambien for a couple of weeks. Blah.

I doubt we would have figured this out on our own, so it's a good thing he made that call. And even the dex will be worth it if it fixes me up.