At least five interesting things for the middle of your week (#29)

AI might help the middle class, Americans can afford food, narratives about science, Europe's defense industry, and Native American YIMBYism

Photo by Alex Haney on Unsplash

Winter is turning into an early spring, with a little helpful push from global warming. The sun is shining, the flowers are blooming, and it’s a great time to be outdoors enjoying the crisp cool air. But you know what’s even better than touching grass? Reading economics blogs, of course!

This week I have only one podcast for you — an episode of Econ 102, where Erik Torenberg and I discuss the curiously persistent delusion among Westerners that Russia represents some sort of attractive or functional alternative to Western civilization:

Anyway, on to our list of interesting things. Today’s theme is “pushback against negative stories and trends”. It’s important to remember that things in the economy are rarely as dire as the pundits would have you believe (unless we’re talking about Europe’s ability to defend itself, in which case things are probably more dire).

1. AI might help restore the middle class, actually

Last week I wrote a gigantic 7000-word review of Power and Progress, a book that claims that AI is going to automate away the middle class. I don’t think the book managed to make its case at all. But still, I can’t rule out the possibility that AI will hollow out the middle class; no one can. It might happen.

But I think there’s reason to be hopeful here. As soon as economists started doing experiments on generative AI, they found an incredibly consistent pattern: AI narrowed the productivity gap between high-skilled and low-skilled workers. I pointed this out in a post last year:

[T]ake this new paper by Brynjolfsson, Li, and Raymond, which measures the effect of an AI tool on the productivity of customer support workers…[T]he AI tool helps lower-skilled workers more than higher-skilled ones…[F]or customer support people who can already do their jobs well, AI provides little or no benefit. But for those who are normally pretty bad at their jobs, or are new on the job, the AI tool boosts their skills immensely…

What’s really interesting is that this isn’t the first study to find this pattern. Noy and Zhang (2023) observe the exact same thing for college students doing writing tasks — the AI uplifts the worst performers a lot more than the top ones. And a team at Microsoft Research found that GitHub Copilot improves the coding performance of older developers and less experienced developers more than others.

I think we’re starting to see a pattern emerge here. Generative AI consistently seems to uplift the lowest performers more than the high flyers…

[R]esults like these make me think it’s possible that generative AI might be the technological “revenge of the normies”. The master craftspeople of software engineering, financial management, and business communication might find that with just a little help from an AI assistant, a bunch of normal people can do their job.

Now David Autor, an economist at MIT who has done some of the key research on automation and inequality, is saying something similar. In a recent article, he argues that computerization helped hollow out the middle class by automating away “middle-skill” jobs, but that AI might reverse this trend:

The unique opportunity that AI offers humanity is to push back against the process started by computerization — to extend the relevance, reach and value of human expertise for a larger set of workers. Because artificial intelligence can weave information and rules with acquired experience to support decision-making, it can enable a larger set of workers equipped with necessary foundational training to perform higher-stakes decision-making tasks currently arrogated to elite experts, such as doctors, lawyers, software engineers and college professors. In essence, AI — used well — can assist with restoring the middle-skill, middle-class heart of the U.S. labor market that has been hollowed out by automation and globalization.

Basically, he makes the same argument I’ve been making about AI as a machine tool for the mind. The machine tools of the industrial age didn’t automate away the jobs of master weavers and smiths; they simply allowed semi-skilled average people to do those jobs just as well as the experts, and at much higher volumes. Replacing high-skilled workers with tech-empowered low-skilled workers is not the same as “automating” work; it’s simply opening it up to the masses.

Autor goes into much more detail than I have, laying out exactly how he thinks AI will complement the mental skills of ordinary people and allow them to compete with the intellectual elite. He uses the analogy of YouTube videos:

If AI can supply cheap expertise by the bucketful, won’t the remaining thimblefuls of human expertise be superfluous? I’ll answer with an analogy: YouTube…Let’s say that I wanted to replace the fuse box in my 19th-century home with a 20th-century circuit breaker panel. Assume, hypothetically, that I’ve never touched a pair of electrical pliers and I don’t own insulated gloves. But I have a free Saturday and there’s a Home Depot around the corner. Confidence high, I fire up one of the dozens to hundreds of YouTube how-to videos on this subject and get to work. Inevitably, but not immediately, I realize that my 19th-century fuse box is not quite like the one in the video. Whether I choose to reverse course or brazenly carry on, I face a palpable risk of a nasty shock or electrical fire…  

To harness the free expertise on tap, I needed foundational expertise: procedural knowledge for handling high voltage circuits, expert judgment for problem-solving when the job went off script. With that expertise in hand, YouTube might have been exactly what I needed.

My point: rather than making expertise unnecessary, tools often make it more valuable by extending its efficacy and scope…

While AI is more than simply YouTube for white-collar professionals, its role in extending the capabilities of experts will be paramount. Most medical procedures, for example, follow a well-specified set of steps. But executing these steps requires hands-on practice and the tacitly acquired expert judgment that comes with it.

Plausibly, an experienced medical worker could, guided by AI, master a new medical device like the use of a new type of catheter, or carry out an unfamiliar procedure in an emergency. An untrained adult might also succeed in catheterizing a patient (or themselves) by following a “how to” video on YouTube. But when that procedure inevitably goes off script, someone with expert medical judgment better be on hand. 

Artificial Intelligence will not in general [automate] high-stakes tasks, like catheterization. But it can enable workers with an appropriate foundation of expertise to level up. AI can extend the reach of expertise by building stories atop a good foundation and sound structure.

In fact, I already see a few encouraging signs of educators shifting what they teach in order to prepare Americans to boost their skills using AI.

Now, skills-based inequality certainly isn’t the only kind. There’s also inequality based on physical capital. Just as the people who owned the machine tools in the Gilded Age pulled away from the rest, the wealth of the AI age might disproportionately flow to Nvidia and TSMC and Microsoft. So if that happens, we’ll have to deal with it, probably via higher capital taxation.

Anyway, Autor’s article is long, but I strongly recommend reading it in full. It’s a useful antidote to the popular doomerism about AI taking away everyone’s job.

2. Yes, Americans can still afford food (they’re just eating out more)

It’s time for another one of those pesky viral charts! This one is courtesy of the Wall Street Journal. In a recent article, the WSJ declares: “It’s Been 30 Years Since Food Ate Up This Much of Your Income”. And they back it up with this startling-looking chart:

Source: WSJ

Wow. Suddenly, Americans can barely afford to eat! That’s horrible!!

Except OK, let’s take a closer look. The first thing we should do is to look at the y-axis. Note that it doesn’t start at zero. A y-axis that doesn’t start at zero will tend to make changes look bigger. If we look at the actual numbers, the WSJ chart shows that spending on food went from a little over 10% of disposable income in the 2010s to a little over 11% now. An increase of 1% of disposable income is more than nothing, but it’s not exactly breaking the bank either.

Next, let’s check the data source. Jeremy Horpedahl notes that the WSJ is using numbers from the USDA, which disagree a little bit with other government data sources. So let’s check the other sources. Here are the numbers from the Bureau of Economic Analysis:

Here we also see food spending — which consists of spending on groceries plus spending on restaurants — increasing by about 1% of disposable income compared to the 2010s. So that checks out. But this chart does two things for us. First of all, the y-axis starts at zero, so you can see how modest the recent increase is. Second, it starts much earlier, so you can see that Americans are currently spending a lot less of their income on food than they did in the 1980s. That is valuable and helpful perspective.

But now let’s take a look at the two types of spending on food: groceries and restaurants. Here’s a chart with grocery spending in blue and restaurant spending in red (both as percentages of disposable income):

As you can see, grocery spending is no higher than it was in the 2010s or the 2000s. It’s restaurant spending that has risen. People are now spending more of their income on going out to eat than ever before. In fact, for the first time ever, restaurant spending is almost equal to grocery spending.

Does a nation of restaurant-goers sound like a nation of people who can’t afford food? To me, it does not.

Now, that doesn’t mean the higher cost of groceries isn’t a burden on Americans. It is. If you go out to eat more but still spend the same percent of your income on groceries, that means you’re a little more burdened than before. But the fact that people are shifting their spending toward going out to eat more suggests that the burden is not as crushing as the WSJ’s viral chart suggests.

Remember, whenever you see a viral chart, make sure you know what you’re looking at.

“Science is getting more expensive” vs. “science is slowing down”

There are basically two “stagnationist” narratives about science. The first is that science is getting more expensive to do — that for any given increase in productivity, we now have to hire a lot more researchers for a lot more hours than before. The second narrative is that scientific discovery is slowing down — that we’re actually discovering fewer new useful things every year than we used to.

These are very different narratives! One says we’re working harder to stay on track, while the other says we’re actually off of the track. Back in 2021, I argued that the first narrative is well-supported by the available evidence, while the second narrative is not. Now I have a bit of new evidence, which strengthens my prior beliefs on both counts.

First, let’s talk about falling research productivity. The seminal paper here is “Are Ideas Getting Harder to Find?”, by Bloom et al. (2020), which you’ve probably seen me reference a million times. It mainly relied on data from the U.S., so it would help to have a more global perspective. But I recently found that Boeing and Hünermund have a 2020 paper called “A global decline in research productivity? Evidence from China and Germany”, showing a very similar slowdown for both Germany and China.

In addition to national-level data, they look at specific companies. In both countries, increased R&D spending at the company level has not been matched by similar increases in output. In other words, German and Chinese companies are also running harder just to stay on track. In China this could be due to the slowing of catch-up growth in general, but Germany is on the productivity frontier, so this probably represents a slowdown in the productivity of pioneering research.

Now on to the second, scarier narrative: the idea that science is actually making fewer discoveries, instead of simply paying more to make each one. In January 2023, Park et al. published a very worrying paper in Nature, entitled “Papers and patents are becoming less disruptive over time”. A lot of people hailed this paper as disturbing proof of a slowdown in science.

I was always highly skeptical of this paper’s methodology; shortly after it came out, I wrote a post explaining some alternative hypotheses to explain their findings. Now it turns out that the problems with Park et al.’s paper are more fundamental. In a new paper, Holst et al. find that Park et al. just made a ton of errors handling their data. Here is Holst et al.’s abstract:

Park et al. reported a decline in the disruptiveness of scientific and technological knowledge over time. Their main finding is based on the computation of CD indices, a measure of disruption in citation networks, across almost 45 million papers and 3.9 million patents. Due to a factual plotting mistake, database entries with zero references were omitted in the CD index distributions, hiding a large number of outliers with a maximum CD index of one, while keeping them in the analysis. Our reanalysis shows that the reported decline in disruptiveness can be attributed to a relative decline of these database entries with zero references. Notably, this was not caught by the robustness checks included in the manuscript. The regression adjustment fails to control for the hidden outliers as they correspond to a discontinuity in the CD index. Proper evaluation of the Monte-Carlo simulations reveals that, because of the preservation of the hidden outliers, even random citation behaviour replicates the observed decline in disruptiveness. Finally, while these papers and patents with supposedly zero references are the hidden drivers of the reported decline, their source documents predominantly do make references, exposing them as pure dataset artefacts.

Basically, Holst et al. find that all of Park et al.’s results are driven by a bunch of papers that apparently don’t reference any previous papers. But when Holst et al. looked closely at those papers, they find that almost all of them actually do reference a bunch of other papers! In other words, Park et al.’s data appears to be complete trash. In fact, disruptiveness, by their chosen definition, doesn’t appear to be declining over time.

So that’s good news. We still don’t have evidence that scientific discoveries are slowing down in the absolute sense.

By the way, if you’re looking for a list of awesome recent scientific discoveries, check out this thread by Eliot Hershberg on recent rapid advances in genetic engineering. As Josh March and Kasia Gora recently wrote on this blog, we really do seem to be off to the races when it comes to bioengineering. The future may be more expensive than it used to be, but it still looks bright.

Can Europe actually make weapons, or only hot air?

French President Emmanuel Macron is suddenly saying a lot of very strong things in support of Ukraine:

French President Emmanuel Macron has said it is key for Europe's security to defeat Russia in Ukraine, amid urgent pleas for more weapons from Kyiv.

He was speaking in Paris where he said that European leaders had agreed to set up a coalition to give Ukraine medium- and long-range missiles and bombs.

He added that there was "no consensus" on sending Western troops to Ukraine, but that "nothing should be excluded".

Germany’s deputy chancellor responded that France should put its money — or more accurately, its weapons manufacturing — where its mouth is:

“I’m pleased that France is thinking about how to increase its support for Ukraine, but if I could give it a word of advice — supply more weapons,” [Deputy Chancellor] Habeck said on Tuesday…[H]e called on France to “do what you can now and give Ukraine the munitions and the tanks that can be supplied now.”

Habeck is right to scold France, which has lagged significantly behind other countries in military aid to Ukraine:

France, with one of Europe's largest military industrial complexes, trails far behind. The Insitute found that French commitments — aid given and promised — were €635 million, while Germany was €17.7 billion, second only to the U.S.

On the other hand, Habeck and the Germans should think about fixing their own deficiencies in this regard. Despite soaring orders for ammunition and weapons, Germany — which most regard as Europe’s industrial heartland — hasn’t managed to increase production by very much:

Source: Joey Politano

This is a complete clown show. America remains divided by internal politics and will have to spend most of its effort facing China soon anyway. European countries, especially Germany and France, are absolutely essential to the defense of Ukraine, and Eastern Europe in general, and ultimately all of Europe against the Russian threat. If these countries combined can’t even manage to build as much ammo as Russia — a nation with a GDP the size of Italy and a population 1/3 that of the EU — then their decline just might be irrevocable.

Germany and France need to step it up in a big way.

Native Americans could help the U.S. build things

Americans have an annoying habit of thinking of our indigenous peoples as something akin to elves in a Tolkien novel — a wise ancient race with a deep, almost mystical connection with the land, with indigenous knowledge that serves as a substitute for modern science. As a result, our programs to return land to Native Americans tend to involve drafting tribal groups into being nature conservationists.

Now, nature conservation isn’t a bad thing, and if tribes want to do that, then fine, no problem. But Canada is pioneering a very different approach toward its First Nations. It’s allowing them, and even helping them, to build modern cities and modern industry.

First, and most famously, there’s the Senakw housing development, a piece of land in downtown Vancouver that the Canadian government is allowing the Squamish Nation to develop as it sees fit. Naturally, the Squamish didn’t return the land to the wild, but instead planned a very solarpunk-looking forest of apartment towers. This has elicited squawks of protest and various political sabotage attempts from local Vancouver NIMBYs, but the political importance of returning land to the First Nations has steamrolled all objections. Dense housing is getting built in a city that badly needs it, and the tribes will reap the economic benefits.

Now in addition to housing, let’s think about manufacturing. The Canadians are helping the Mahalat Nation to build the country’s largest battery factory on Vancouver Island:

Vancouver Island is set to see one of its largest high-tech industrial manufacturing investments in decades, and it is also a project owned and spearheaded by the Malahat First Nation.

It was recently announced that the First Nation has partnered with Energy Plug Technologies to build a 100,000 sq ft gigafactory for manufacturing and assembling lithium-iron-phosphate battery storage products…This is a partnership with Vancouver-based Energy Plug Technologies, which will have a 49% stake in the facility, with the First Nation holding the majority ownership interest of 51%…These grid-scale battery storage systems will be suited for residential, commercial, and industrial uses, with the products ranging from 5 kW to 50 MW. With the acceleration in the transition from fossil fuel-powered building systems to electrification, there is growing global demand for such products.

In other words, the First Nations are helping Canada to do industrial policy!

This could be done in the United States as well. Just as Native American tribes saw an influx of money from casinos, they could see an even bigger influx of money from battery factories and solar farms on their land. A push for strong legal protections for Native American land rights — perhaps spearheaded by Supreme Court Justice Neil Gorsuch — would allow the tribes to overrule local NIMBYs. At a time when NIMBYism and permitting hangups are threatening America’s push for green industrial policy, a win-win alliance with Native American tribes could be exactly what this nation needs.

Instead of demanding Native Americans be guardians of a pastoral past, we could enlist their aid in leading us into a green technological future.

---