• The Robots Are Coming! But Not For a While.

    Things are slow this morning, so I’m going to rail against one of my pet peeves: analogies to the Industrial Revolution as evidence that robots won’t reduce employment. My victim today is Heather Long at Wonkblog:

    History suggests new jobs will replace old ones. As the Industrial Revolution demonstrated, technological transformations create new jobs no one has thought of yet. The same trend appears to be happening today.

    Companies shed workers during the Great Recession and rapidly tried to cut costs, including by introducing more machines on assembly lines and in fast casual restaurants like Panera, where you can now order on a touch screen. Yet even with those trends, the U.S. economy has added more than 16.4 million jobs since the low point for employment in December 2009.

    “Tom Cotton is woefully misinformed,” said Joe Brusuelas, chief economist at RSM. “Robots will create more jobs.” Brusuelas points out that many of the fastest-growing jobs today, such as “user design” and “cloud engineers,” weren’t around a decade ago. We like to talk about how robots kill jobs, but we tend to talk a lot less about how many other jobs are being created in the economy.

    Ah yes: user design and cloud engineers. All those displaced truck drivers will just transition into Silicon Valley engineering jobs. Sure they will.

    We have had one (1) “technological transformation” in recent human history. It did, eventually, create new jobs for the class of workers¹ who had been displaced. However, this sample size of one provides no evidence that all technological transformations will work the same way.

    In fact, it’s vanishingly unlikely that the AI Revolution will work the same way as the Industrial Revolution. The latter displaced human muscle power, but all those machines still had to be run by humans with brains. That’s where the new jobs came from. The AI Revolution will displace human brainpower, and that means there will be nothing left for most humans to do. If a robot is both mentally and physically equal (or superior) to a human being, then by definition they can do whatever a human can do. We don’t have to argue one by one about every job category, or even think about every possible new job that might be created in their place. The answer to all of them is: robots will do that too.

    Now, if you don’t think AI will ever get to human level, that’s fine. Then the AI Revolution will never happen. I’d still argue that enough of it will happen to cause considerable upheaval, but now we’re in a technological argument. What happens, for example, if AI and robotics get smart enough to perform all unskilled and semiskilled labor, but never advance further than that? We won’t have 100 percent unemployment, but we’ll have 30-40 percent unemployment, and that strikes me as still a big enough problem to worry about.

    However, if you accept that AI will eventually get to human level, then yes, robots will take all our jobs. And they’ll take all the new jobs too.

    As for Cotton, though, he’s also wrong. There’s no conflict between the American economy needing more workers in the short term but being in danger of having too many workers in the longer term. Today we’re short of workers. In 20 years we’ll be wondering what to do with all the workers we have. There’s no reason to think those two things are in tension.

    ¹I say “class of workers” because many of the actual workers displaced by the Industrial Revolution were indeed put out of jobs. It was only later that other jobs cropped up to replace them. Workers of similar skill levels then took those jobs, but that was no help to the original folks who had lost their livelihoods.

  • Are “Deaths of Despair” Related to Lead Poisoning?

    Here’s something I want to toss out, even though I don’t have an answer to offer—or even a clue, really. It’s about lead.

    People periodically ask me about whether lead might be responsible for some phenomenon or other. I’m generally very careful about this stuff. It’s possible that lead poisoning is responsible for lots of things, but in most cases the effect is likely to be (a) fairly small and (b) swamped by other factors anyway. That makes it all but impossible to measure, and if you can’t measure it you can’t really say anything about it.

    The things that are easiest to measure are behaviors at the far end of the bell curve. Here’s why: if lead affects some behavior by a little bit, it will push the mean of the bell curve over by a point or two. Most of the time, this is just too small to measure. But the tail of the bell curve might be doubled or tripled in size. This is the case with violent crime, for example, which is why it’s feasible to correlate lead and crime.

    What other behaviors are fairly rare, and therefore might increase by a large amount due to lead poisoning? For some reason, it recently occurred to me to be intrigued by the Case-Deaton study of “deaths of despair.” There are some technical problems with their original paper, but I think everyone agrees that even when those are corrected there’s still something going on. And that something is primarily affecting people who are about 50 years old.

    The peak years for lead poisoning in the US were 1965-1975. People born in those years are now ages 43-53. Is there some kind of connection?

    The fact that this doesn’t seem to be happening in other countries—which is pretty much the whole point of the Case-Deaton paper—suggests that lead may not be the culprit. There’s also the fact that it seems to affect whites more than blacks, which is the opposite of what you’d expect if lead was involved. At the same time, it’s possible that lead has an effect that only shows up if other conditions help it along. Who knows?

    In any case, it seems like this might be worth a look. I’m not entirely sure how to look at it, though. Take tooth samples of folks who drink themselves to death to check for lead concentrations? That’s not especially likely. But maybe there are less intrusive ways of establishing whether there’s a correlation in the first place. If there is, it would be worth further study.

  • “Don’t Feed the Trolls” Is Even Better Advice Than it Used to Be

    You should think long and hard before sharing this tweet with the world. Unfortunately, the people who most need this advice are the least likely to heed it.

    Over at the Atlantic, Conor Friedersdorf calls out lefty call-out culture. It’s become so excessive, he says, that Twitter mobs routinely go ballistic over the smallest, most inadvertent micro-slights, even those from folks who are basically on their side:

    I don’t understand why they believe that extreme anger and stigma should be directed at people whose intentions and substantive beliefs are so close to their own….

    I don’t understand why they dedicate so much energy and focus to what even they call microaggressions at a time when an ascendant coalition in American politics is bent on deporting as many immigrants as possible….

    I don’t understand how they think they can defeat that nativist faction if their own pro-immigrant coalition engages in divisive infighting over transgressions as inevitable as clumsy wording….

    Even if every object of dragging deserved it, I don’t understand how the outcome could be anything other than punishing an infinitesimal percentage of bad actors while turning off so many with the excesses that it provokes a backlash.

    Over at Vox, I think Ezra Klein coincidentally provides most of the answer in an interview with Tristan Harris:

    Ezra Klein: I had Jaron Lanier on this podcast a couple months ago, and he said something I’ve been thinking about since then. He said that the key to a lot of social media is [that] negative emotions engage more powerfully than positive emotions. Do you think he’s right about that?

    Tristan Harris: Oh, absolutely. Outrage just spreads faster than something that’s not outrage. When you open up the blue Facebook icon, you’re activating the AI, which tries to figure out the perfect thing it can show you that’ll engage you. It doesn’t have any intelligence, except figuring out what gets the most clicks. The outrage stuff gets the most clicks, so it puts that at the top….If the first thing you do when your eyes open is see Twitter and there’s a bunch of stuff to be outraged about, that’s going to do something to you on an animal level.

    Journalists as a group evaluate social media poorly, and we evaluate Twitter especially poorly. Think about how Twitter works. There are a very few influencers who are determined to root out and denounce anything that’s even remotely problematic. They do this mostly via absurdly hostile readings of other tweets or by making connections that most people would never notice. Nonetheless, once that bell is rung, it can’t be unrung—and their followers all rush in to denounce the micro-slight in question. Why do the influencers do this? Because they’re zealots, and that’s what zealots do. And why do they attract mobs who follow them so uncritically? Because those are the kinds of mobs zealots always attract.

    It’s exhausting to be on the receiving end of this stuff, but it’s truly meaningless. There will always be zealots and their mobs looking for outrages to slay. And while Twitter makes them more visible, their numbers are still tiny. A few hundred? A few thousand? That’s nothing considering the minuscule effort it takes to dash off a bit of tweetrage. Unless a Twitter mob gets into the 10-100,000 range, it simply doesn’t represent anything important.

    Even among the far reaches of the left, I imagine that most people agree with Friedersdorf that outrage is a stupid response to micro-slights. So the answer to his bewilderment, I think, is twofold. First, social media is a magnet for outrage, and the platforms themselves encourage this because it keeps people engaged and delivers more eyeballs to their advertisers. Second, even given this, the number of people outraged by micro-slights is truly insignificant. Social media tidal waves, in which a few thousand responses rain down within a couple of hours, merely make them seem big.

    If you ignore small Twitter mobs—and by small, I mean at least anything under 10,000 tweets—most of the paradoxes and conundrums of the social justice zealots go away almost instantly.

  • Republicans Are Trying Out a Shiny New Excuse For the Great Kansas Failure

    Last week a reader emailed me about a new meme he had just come across:

    Heard a random Republican talking head on NPR recently, and when the interviewer questioned him on the “Kansas experiment,” his automatic response was a) Kansas “massively” increased spending when they cut taxes, so that’s why they have problems; and b) North Carolina has done the same thing without the increased spending and it’s working great.

    Of course this smells like bullshit to me, but I don’t actually know. Are either of these assertions correct?

    I’m too lazy to waste time on North Carolina right now, but spending in Kansas is easy enough to check. Here it is:

    Since 2011, when Sam Brownback took office promising a “red state experiment,” general fund spending has been flat while spending from all sources has declined by 1.7 percent. I’m pretty sure this doesn’t count as “massively” increasing spending.

    Bottom line: Brownback slashed taxes, kept spending flat, wrecked Kansas schools, and turned in lousy economic performance compared to his neighboring states:

    By just about any measure, the red-state experiment failed, and Republicans can hardly run away from Kansas fast enough. I guess their latest wheeze is to pretend that Brownback was a faker all along and it was really North Carolina we should have kept an eye on. Uh huh.

  • Lunchtime Photo

    Ansel Adams became famous for making iconic pictures of Yosemite National Park like “Moon and Half Dome.” I too seek photographic immortality, so I’m following in his footsteps. Ladies and gentlemen, today I present “Airplane and El Capitan.”

  • Pennsylvania Gets a New Map

    Via the Sunbury Daily Item, here is the new congressional map for Pennsylvania:

    Republicans flatly refused to create a non-gerrymandered map, so the Pennsylvania Supreme Court created this one for them. It’s the map that will be used for this year’s midterm elections unless Republicans are able to get a federal court to issue an injunction of some kind. Since the Supreme Court’s decision was based entirely on state law, Republicans seem unlikely to succeed, but you never know. If you find the right judge….

  • Why Do Women Earn Less Than Men?

    Sarah Kliff points today to a new study from Denmark on the gender wage gap. Danes are famously egalitarian, and labor force participation is nearly equal between men and women these days. However, Denmark still has a large gender wage gap—nearly as large as the United States, in fact. Why? Researchers Henrik Kleven, Camille Landais, and Jakob Egholt Søgaard conclude that it’s almost purely a childbearing penalty:

    I’ve overlaid men and women in a single chart here to make it easier to compare them. Earnings are shown relative to the year before the birth of a first child, and the trajectories of men and women are similar prior to childbirth. Generally speaking, men’s earnings start a little higher but women’s earnings accelerate a little faster. For women without children and for all men, regardless of whether they have children, these trajectories continue throughout their careers: men suffer no earnings penalty at all when they have children. However, women who have children take a huge hit in earnings.

    This is unsurprising. The big question is why they have lower earnings. Here’s a panel of charts that breaks it down:

    After childbirth, fewer women work; they work fewer hours; and they get lower wages. And this is unrelated to education level: college graduates bear childbirth penalties that are about the same as high school grads. In fact, nearly all gender inequality has been wiped out in Denmark except for the gender gap due to childbirth:

    However, this still doesn’t answer the question of why. Do women with children work less out of preference, or because firms treat them badly and eventually some of them give up? There’s a limit to what administrative data can tell us, but by expanding their dataset the authors are able to conclude that some of it is due to family influence:

    Women incur smaller earnings penalties due to children if they themselves grew up in a family where the mother worked more relative to the father….The size of this effect is roughly unaffected by including the detailed non-parametric controls for education and wealth….[This] suggests that female child penalties are driven partly by female preferences formed during her childhood, rather than by male preferences formed during his childhood.

    Women from more traditional families form an early preference for working less when they have young children to take care of. Women from more liberal families don’t. In other words, it’s women from traditional families who account for the biggest share of the childbearing penalty. However, the size of the difference between traditional and liberal families isn’t large, so there’s clearly a lot more going on than just that.

    That’s the case in Denmark, anyway. Is something similar true in the United States? We lack the detailed administrative data of Denmark, so it’s not easy to conduct a similar study here. However, American studies do show that the gender gap in earnings opens up mostly between ages 25 and 35, which certainly suggests that children are the prime cause. More research, please.

  • Donald Trump Is America’s Worst, 2nd Worst, or 5th Worst President

    Today is Presidents/President’s/Presidents’ Day, and the New York Times is celebrating with a bit of clickbait that ranks all 44 presidents.¹ It turns out that presidential scholars of the left outrageously rank Donald Trump as our worst president after only a year in office, which means we have to turn to right-leaning scholars to get a more considered view of things:²

    Number 40! Now that’s more like it. Trump is better than the guy who sleepwalked into the Civil War and the general who died after only 40 days. But he’s still worse than the Teapot Dome guy and the guy who inspired the Mallard Fillmore comic strip. Sad.

    ¹In presidential numbering, Grover Cleveland counts twice since he had two separate terms, so we’ve had 45 presidents. In presidential rankings, Cleveland is just one person, so we’ve had 44 presidents. Got it?

    ²It’s worth noting that presidential scholars, just like us common folks, seem to have a bias for the present. Six out of 14 postwar presidents make their top 10.

  • We Should Ban Semi-Automatic Firearms

    Beretta

    I’m not generally on the gun control beat, but I’ll repeat my view for the record: semi-automatic weapons should be banned for civilian use. Basically, shotguns, revolvers, and bolt-action rifles would remain legal, and that’s it.

    The last time I mentioned this, a bunch of gun folks chimed in to claim I was an idiot. Revolvers are semi-automatic weapons! Ha ha ha. Being the reasonable guy that I am, I was willing to consider them manual load weapons, since it takes human power to advance the cylinder. A true semi-automatic uses the power of expanding gas¹ to chamber a new load. However, if the gun folks consider a double-action revolver to be a semi-automatic, who am I to argue? That just means my list has been revised to include shotguns, single-action revolvers, and bolt action rifles.² In other words, the only legal firearms would be those that require a separate human action to load a new round. There are other details I’d support too, but this is the main thing.

    If I were your benevolent dictator, this is what would would happen. But I’m not, and nothing like this will happen in my lifetime. This is why I don’t spend much time writing about guns.

    ¹Or, in the case of most handguns, recoil force.

    ²Or pump action or lever action. You get the idea, I’m sure.

  • How Many Threats Can the FBI Evaluate on a Daily Basis?

    FBI Director Christopher Wray.Ron Sachs/CNP via ZUMA

    The FBI has taken a lot of criticism for failing to follow up on a warning about the teenager who killed 17 schoolchildren in Florida last Wednesday. Here’s the BBC’s report:

    On 5 January a person close to the teenager contacted the FBI tipline to provide “information about Cruz’s gun ownership, desire to kill people, erratic behaviour, and disturbing social media posts, as well as the potential of him conducting a school shooting”, said an FBI press release….In 2016, the FBI received about 1,300 tips a day through its website, which is staffed around the clock by two dozen people. In addition to online tips, FBI field offices receive dozens of calls. About 100 of the tips are considered “actionable”.

    This means that in January the FBI received something on the order of 50,000 tips. If they spend an average of, say, an hour on each one, that’s about 300 agents working full time doing nothing but investigating tips. Or, perhaps it means a thousand agents spending a quarter of their time on tips. Are they staffed to do that? What exactly is the protocol for responding to this tidal wave of tips?

    But even that isn’t the real question. Suppose they had investigated Cruz more thoroughly. What could they have done? It’s not illegal to own a bunch of high-powered guns. It’s not illegal to rant on Twitter or Facebook. The FBI could have interviewed the guy, but unless he’s broken the law that’s about the end of it. It’s not clear to me what the FBI could have done here even if they had followed every protocol to the letter.

    I’m genuinely curious about this. How well staffed is the FBI to handle tips? What can they do against a motivated attacker aside from an interview? To the extent that conservatives are using this failure as a handy excuse to attack FBI Director Christopher Wray, I don’t care what happened. These scattershot attempts to suck up to Donald Trump are too patently phony to worry about. But to the extent there might truly be something we can do better, I do care. Has anyone made any concrete suggestions on this score?