At least five things to finish your week (#19)
Real wages, why North Korea is poor, China's stalled catch-up, the economics of social media, AI takes some jobs, and a comic about rabbits
Howdy, folks! I am back from my travels in Ireland and Singapore, and there’s lots to catch up on — especially because I’ve been blogging a little too much about geopolitics lately and letting a lot of interesting econ stuff go by. Anyway, there’s lots of fun stuff in this week’s roundup.
But first, podcast! Episode 19 of Econ 102 with Erik Torenberg is here:
Anyway, on to this week’s At Least Five Interesting Things, of which there are actually six:
1. Are real wages up or down?
Over the next year, the big economic debate is going to be whether Americans are better off under Joe Biden; this is just how election years work. And one of those debates will be about whether wages have gone up or down, in real terms — i.e., whether an hour of work is able to afford Americans more eggs, gasoline, rent, health care, etc. than before Biden took office.
The answer is complicated. First of all, it depends on your comparison point. Real hourly compensation (which includes wages and benefits) is:
higher than before the pandemic,
below the pre-pandemic trend line,
down from early 2021 when Biden took office, and
up from mid-2022.
Here’s a graph of those four comparisons:
Obviously, which of these comparisons you think is more important will depend on
the last video you saw on TikTok your political point of view and your personal economic situation.
But in fact, things get even more complicated for two reasons:
Wages for the typical worker are different than average wages, because very high wages for people at the top (CEOs and AI engineers and such) will skew the average number a lot.
The pandemic resulted in low-wage workers being laid off at disproportionate rates, which mechanically raised the average much more than wages rose for the people who kept their jobs.
Labor economist Arin Dube puts the first of these in a chart, showing us that hourly earnings for production and nonsupervisory workers have done a lot better than the overall average:
(Note: hourly earnings are somewhat different from hourly compensation, since compensation includes benefits like health care payments and retirement contributions.)
As for the composition effect, it’s much harder to assess. Joey Politano has a good blog post going through various wage measures, and he notes that when you hold the composition of jobs constant, wages have actually done worse than the headline numbers. Here’s a chart of the real Employment Cost Index, which adjusts for job composition, compared to real hourly compensation:
But as Politano points out, this can be deceptive too; in the years since the pandemic, a lot of Americans have been switching out of lower-wage jobs and into higher-wage jobs, which is good, and which the Employment Cost Index will totally miss.
So overall we have a mixed picture here. Real wages are growing again, which is good, and they’re growing even faster for people in the middle and bottom of the distribution, which is also good. But the big hit they took in 2021 and 2022 hasn’t been fully erased.
2. Why is North Korea so poor?
I found — quite by accident — that one easy way to make leftists on Twitter angry is to post an image of the Korean peninsula at night. The famous one at the top of this post is pretty old now, but there are plenty of others, and yes, it still looks like that. North Korea is an extremely poor country, while South Korea is an extremely rich one. This fact is often used as proof that communism is bad and capitalism is good.
But why exactly is North Korea so poor? Cuba and the USSR were both communist, and they were much less poor than North Korea. And in recent years, nominally “communist” regimes like China and Vietnam have experienced rapid development, while North Korea has not. What’s the difference?
One explanation, beloved of leftists, is that the U.S. simply bombed North Korea back to the stone age 70 year ago, and it remained there. This isn’t credible. Even if you believe the bombing of North Korea was more destructive than the devastation of South Korea — whose capital changed hands three times during the war — it still doesn’t fit with North Korea’s timeline. In fact, until the early 1970s, North Korea was growing at a decent clip, and keeping pace with South Korea.
Obviously these figures are hard to know exactly, and the perfectly straight line for North Korea between 1973 and 1992 just means “we don’t have any data for this period”. But in 1965, the economist Joan Robinson was praising North Korea’s rapid economic growth and successful rebuilding efforts:
Eleven years ago in Pyonygang there was not one stone standing upon another. (They reckon that one bomb, of a ton or more, was dropped per head of population.) Now a modern city of a million inhabitants stands on two sides of the wide river, with broad tree-lined streets of five-story blocks, public buildings, a stadium, theaters (one underground surviving from the war) and a super-de luxe hotel. The industrial sector comprises a number of up-to-date textile mills and a textile machinery plant. The wide sweep of the river and little tree-clad hills preserved as parks provide agreeable vistas. There are some patches of small gray and white houses hastily built from rubble, but even there the lanes are clean, and light and water are laid on. A city without slums.
So what happened in the 70s? Well, for one thing, North Korea embarked on a massive military buildup in order to reduce its military dependence on Russia and China. This diverted resources from investment in the civilian economy. North Korea also borrowed a bunch of money internationally and invested heavily in mining; when it was hit hard by a long decline in the price of metals in the 70s, 80s, and 90s, it was unable to pay its foreign debts and had to default. That, along with the collapse of the USSR in 1991 — which had provided North Korea with large amounts of aid and trade — provided a long-running macroeconomic shock that took a lot of time to recover from.
That was, of course, on top of the typical communist mismanagement of the microeconomy. North Korea has always been one of the world’s most centrally planned economies; extreme central planning can be good for conducting a war or rebuilding from a war, but otherwise tends to misallocate resources by quite a lot. There is some optimal level of central planning, and North Korea simply exceeded it. The principle of “juche” (self-reliance) also discouraged North Korea from both exporting manufactured products — which was one of the main ways South Korea improved its productivity and got rich — and importing goods that could have been used as inputs into domestic production.
Western sanctions have also been a factor in North Korea’s economic isolation, but probably less than people think. Most sanctions are very recent, and are due to the country’s nuclear program; before that, North Korea actually exported a lot of minerals. Meanwhile, China doesn’t impose sanctions and in fact trades with North Korea a fair amount, though thanks to China’s mercantilist policies and North Korea’s general industrial backwardness, North Korea runs a large trade deficit and mostly just sells China minerals. Meanwhile, North Korea has generally resisted efforts by South Korea to expand economic ties.
In other words, North Korea’s poverty is really not hard to explain. When you’re a small country and you close yourself off to trade and you intentionally mismanage your domestic economy, you’re going to be poor. If the U.S. could reach out to North Korea and help ease it out of its paranoia, the situation might change. That’s unlikely, but we did just do it with Vietnam, so maybe there’s hope.
3. Is China really falling behind the U.S. economically?
There’s a fair amount of American economic triumphalism surrounding the APEC summit. Broadly I think there’s some justification for this; the U.S. economy is doing great, while China’s is in the dumps. But a lot of people are sharing this chart of Chinese GDP relative to U.S. GDP as evidence that China’s economic catch-up has halted:
China has certainly had a long-term growth slowdown since the early 2010s, but this chart overstates it. If you read the fine print, you’ll notice that this data is at market exchange rates, which means this measure of GDP depends not just on how much a country produces, but on how cheap or expensive its currency is. And China’s currency, the RMB (typically called the yuan), has, with a few brief interruptions, been getting cheaper and cheaper relative to the U.S. dollar for the past decade:
Of course, some of the reason for the depreciation of the yuan is based on lower expectations of future growth. But some of this is probably due to China’s policy of keeping the yuan from appreciating a bunch in the bullish times (which it does by having its central bank buy U.S. bonds). And some of it is a macroeconomic response to Trump’s tariffs in late 2018; tariffs tend to push down the exchange rate of the country subject to the tax, thus offsetting some of the effects of the trade barrier.
And remember that a cheap yuan makes China’s exports more competitive vs. the exports of the United States. Thus, a combination of macroeconomic weakness and exchange rate management is helping China gain and hold market share in all the industries that the U.S., Europe, and the rest of Asia would like to build up. In other words, part of what you’re looking at when you see that graph above is a story about slowing Chinese growth, but part of it is about increasing Chinese competitiveness. It would be helpful to remember that.
4. Is social media a trap?
In general, economists tend to think that when people can buy more stuff, it makes them more satisfied — after all, if you didn’t want the stuff, you would just not buy it, or throw it away, right? But in the real world, the enjoyment you derive from buying or not buying something often depends on whether other people are buying it. This is called an “externality”.
Sometimes these externalities are positive. For example, using AOL Instant Messenger by myself in the 2000s wouldn’t have been very fun. But when my friends got on it, it became much more useful. That’s a positive network externality, and it applies to most social media.
But sometimes the externalities are negative. An example that’s not often discussed is FOMO, i.e. fear of missing out. If I’m 10 years old and a new video game comes out and all my friends get it, but my parents won’t buy it for me because they think video games are a waste of time and money, I’m immediately less happy than I was before. (Not that this example comes from bitter personal experience, mind you.)
This FOMO is probably particularly intense for social networks. If I wanted to invite my friends to my house in 2003, they would probably want to do so as well, since that was the best way for them to get social interaction. But fast-forward to 2023, and they’ll probably be more reluctant to do so, because they get a lot more social interaction on Instagram, Snapchat, Twitter, etc. (Surveys consistently show that young people are spending less time with friends than before.) So if I’m someone who would rather hang out in real life than on a phone, the new era is worse for me than the old era.
In fact, the new era can be worse for the population overall. Maybe everyone would prefer a general ban on social media, so that they wouldn’t feel the constant FOMO of wondering what was happening on Instagram or worry about having to know the latest TikTok craze. But since the products exist, it might be that consumers in general are just trapped there by each consumer’s individual FOMO.
[T]he authors design a largescale, online experiment aimed at measuring consumer welfare in the presence of both network effects—the phenomenon wherein the value of joining vs. not joining increases with the number of consumers—and consumption spillovers to non-users—for example, fear of missing out on the latest TikTok trend. Their survey-based experiment focuses on TikTok and Instagram and is administered to 1,000 college students.
They find the following:
Users would need to be paid $59 to deactivate TikTok and $47 to deactivate Instagram if others in their network were to continue using their accounts.
Users would be willing to pay $28 and $10 to have others, including themselves, deactivate TikTok and Instagram, respectively…64% of active TikTok users and 48% of active Instagram users experience negative welfare from the products’ existence. Participants who do not have accounts would be willing to pay $67 and $39 to have others deactivate their TikTok and Instagram accounts, respectively.
Taken together, these results imply the existence of a “social media trap” for a large share of consumers, whose utility from the platforms is negative but would have been even more negative if they didn’t use social media.
In other words, the authors think the world might be better off if we just banned social media. But what this experiment doesn’t test is whether some other product would just emerge to create similar amounts of FOMO. Maybe social media really is uniquely bad because of the network effect, or maybe human societies simply have an innate tendency to create must-have goods.
In any case, it’s disquieting food for thought, especially in light of the other potential harms of smartphone-enabled social media.
5. It looks like AI took some jobs
In a post about AI and jobs back in April, I wrote that the idea of a job being “automated” isn’t even well-defined:
What the heck does it mean for a job to be automated?
Does it mean that a human is replaced by a machine and goes onto the welfare rolls?
Does it mean that a human is replaced by a machine and goes and gets a similar job for a different employer?
Does it mean that a human is replaced by a machine and goes and moves to a different kind of job with the same employer?
Does it mean that a human is replaced by a machine and goes and gets a different kind of job with a different employer?
Does it mean that a human uses a machine to do some of her job tasks and does less work overall, while retaining the same job title with the same employer?
Does it mean that a human uses a machine to do some of her job tasks, while taking on additional new job tasks, and retaining the same job title with the same employer?
Does it mean that a human uses a machine to do some of her job tasks, while taking on additional new job tasks, and changing her job title while remaining with the same employer?
…and so on.
Anyway, we now have a pretty concrete example of a job being automated by AI. A new working paper by Hui, Reshef, and Zhou finds that when ChatGPT was released, demand for freelancers on an online freelancing platform immediately went down. Here’s a graph by the ever-excellent John Burn-Murdoch:
This result is pretty clean; the methodology is simple and obvious, and the causality is clear. That said, there are a couple of questions we need to answer before we really know the impact of ChatGPT on the jobs of freelancers:
Did the freelancers get other jobs elsewhere, or become unemployed?
If they got jobs elsewhere, were those jobs in similar fields (e.g. checking the output of AIs for hallucinations, etc.)?
If they got jobs elsewhere, did those jobs pay more or less than their previous freelancing gigs paid?
In general, any new technology will shift the demand for the types of jobs people do, destroying demand in some occupations and boosting it in others. The balance of these is very hard to measure, especially over the long run.
It’s possible that many of the freelancers who lost gigs from the emergence of ChatGPT will gain better jobs elsewhere — for example, I was a freelance editor in Japan before I went to grad school, but I make more money as a result off getting out of the freelance world and getting my PhD. Maybe if freelancing hadn’t been so lucrative at the time I would have gone to grad school a year earlier. (Whether that would have made me happier is another question entirely.)
But anyway, this result pretty clearly shows one instance of occupational demand destruction by new AI technology.
6. A comic about rabbits
Back in June, after my rabbit Cinnamon died, I wrote a post explaining why I like taking care of rabbits. Well, my friend Christine Villanueva, an illustrator and cartoonist (and fellow rabbit lover), took some of the words I wrote in that post and made some adorable little comics out of them:
Here are three of the panels:
Anyway, check out the whole thing!