Month: November 2024
Tariff sentences to ponder
In a September 2024 report, UBS, an investment banker, predicted both tech hardware and semiconductors to be among the top four sectors that would be hardest hit by a general tariff. Their analysis is spot on. Many of the hardware components that make AI and digital tech possible rely on imported materials not found or manufactured in the United States. Neither arsenic nor gallium arsenide, used to manufacture a range of chip components, have been produced in the United States since 1985. Legally, arsenic derived compounds are a hazardous material, and their manufacture is thus restricted under the Clean Air Act. Cobalt, meanwhile, is produced by only one mine in the U.S. (80 percent of all cobalt is produced in China). While general tariffs carry the well-meaning intent of catalyzing and supporting domestic manufacturing, in many critical instances involving minerals, that isn’t possible, due to existing regulations and limited supply. Many key materials for AI manufacture must be imported, and tariffs on those imports will simply act as a sustained squeeze on the tech sector’s profit margins.
Info Finance
Excellent post by Vitalik on prediction markets and the broader category of what he calls info finance:
Now, we get to the important part: predicting the election is just the first app. The broader concept is that you can use finance as a way to align incentives in order to provide viewers with valuable information.
…Similar to the concept of correct-by-construction in software engineering, info finance is a discipline where you (i) start from a fact that you want to know, and then (ii) deliberately design a market to optimally elicit that information from market participants.
One example of this is prediction markets: you want to know a specific fact that will take place in the future, and so you set up a market for people to bet on that fact. Another example is decision markets: you want to know whether decision A or decision B will produce a better outcome according to some metric M. To achieve this, you set up conditional markets: you ask people to bet on (i) which decision will be chosen, (ii) value of M if decision A is chosen, otherwise zero, (iii) value of M if decision B is chosen, otherwise zero. Given these three variables, you can figure out if the market thinks decision A or decision B is more bullish for the value of M. 
Info finance as a three-sided market: bettors make predictions, readers read predictions. The market outputs predictions about the future as a public good (because that’s what it was designed to do).
Importantly, Vitalik notes that AI agents can make decision and prediction markets more liquid at much lower cost.
One technology that I expect will turbocharge info finance in the next decade is AI (whether LLMs or some future technology). This is because many of the most interesting applications of info finance are on “micro” questions: millions of mini-markets for decisions that individually have relatively low consequence. In practice, markets with low volume often do not work effectively: it does not make sense for a sophisticated participant to spend the time to make a detailed analysis just for the sake of a few hundred dollars of profit, and many have even argued that without subsidies such markets won’t work at all because on all but the most large and sensational questions, there are not enough naive traders for sophisticated traders to take profit from. AI changes that equation completely, and means that we could potentially get reasonably high-quality info elicited even on markets with $10 of volume. Even if subsidies are required, the size of the subsidy per question becomes extremely affordable.
Difficult to pronounce names
We test for labor market discrimination based on an understudied characteristic: name fluency. Analysis of recent economics PhD job candidates indicates that name difficulty is negatively related to the probability of landing an academic or tenure-track position and research productivity of initial institutional placement. Discrimination due to name fluency is also found using experimental data from prior audit studies. Within samples of African Americans (Bertrand and Mullainathan 2004) and ethnic immigrants (Oreopoulos 2011), job applicants with less fluent names experience lower callback rates, and name complexity explains roughly between 10 and 50 percent of ethnic name penalties. The results are primarily driven by candidates with weaker résumés, suggesting that cognitive biases may contribute to the penalty of having a difficult-to-pronounce name.
That is from a new AEJ piece by Qi Ge and Stephen Wu.
What do unions do?
This paper shows that immigration fostered the emergence of organized labor in the United States. I digitize archival data to construct the first county-level dataset on historical U.S. union membership and use a shift-share instrument to isolate a plausibly exogenous shock to the labor supply induced by immigration, between 1900 and 1920. Counties with higher immigration experienced an increase in the probability of having labor unions, the number of union branches, the share of unionized workers, and the number of union members per branch. This increase occurred more prominently among skilled workers, particularly in counties more exposed to labor competition from immigrants, and in areas with less favorable attitudes towards immigration. Taken together, these results are consistent with existing workers forming and joining labor unions for economic as well as social motivations. The findings highlight a novel driver of unionization in the early 20th-century United States: in the absence of immigration, the average share of unionized workers during this period would have been 22% lower. The results also identify an unexplored consequence of immigration: the development of institutions aimed at protecting workers’ status in the labor market, with effects that continue into the present.
That is from a new paper by Carlo Medici of Brown University. Via the excellent Kevin Lewis.
Saturday assorted links
China’s Libertarian Medical City
You’ve likely heard of Prospera, the private city in Honduras established under the ZEDE (Zone for Employment and Economic Development) law, which has drawn global investment for medical innovation. The current Honduran government is trying to break its contracts and evict Prospera from 
Honduras. The libertarian concept of an autonomous medical hub, free to attract top talent, pharmaceuticals, medical devices, ideas, and technology from around the world is, however, gaining traction elsewhere—most notably and perhaps surprisngly in the Boao Hope Lecheng Medical Tourism Pilot Zone in Hainan, China.
Boao Hope City is a special medical zone supported by the local and national governments. Treatments in Boao Hope City do not have to be approved by the Chinese medical authorities as Boao Hope City is following the peer approval model I have long argued for:
Daxue: Medical institutions within the zone can import and use pharmaceuticals and medical devices already available in other countries as clinically urgent items before obtaining approval in China. This allows domestic patients to access innovative treatments without the need to travel abroad…. The medical products to be used in the pilot zone must possess a CE mark, an FDA license, or PMDA approval, which respectively indicate that they have been approved in the European Union, the US, and Japan for their safe and effective use.
Moreover, evidence on the new drugs and devices used within the zone can be used to support approval from the Chinese FDA–this seems to work similar to Bartley Madden’s dual track procedure.
Daxue: Since 2020, the National Medical Products Administration has introduced regulations on real-world evidence (RWE), with the pilot zone being the exclusive RWE pilot in China. This means that clinical data from licensed items used within the zone can be transformed into RWE for registration and approval in China. Consequently, medical institutions in the zone possess added leverage in negotiations with international pharmaceutical and medical device manufacturers seeking to enter the Chinese market.
… This process significantly reduces the time required for approval to just a few months, saving businesses three to five years compared to traditional registration methods. As of March 2024, 30 medical devices and drugs have been through this process, among which 13 have obtained approval for being sold in China.
The zone also uses peer-approval for imports of health food, has eliminated tariffs on imported drugs and devices and waived visa requirements for many medical tourists.
To be sure, it’s difficult to find information about Boao Hope medical zone beyond some news reports and press releases so take everything with a grain of salt. Nevertheless, the free city model is catching on. There are already 29 hospitals in the zone including international hospitals and hundreds of thousands of medical tourists a year. The medical zone is part of a larger free port project.
Prospera is ideally placed for a medical zone for North and South America. The Honduran government should look to China’s Boao Hope Medical Zone to see what Prospera could achieve for Honduras with support instead of oppositon.
Hat tip: MvH.
Human Capital Accumulation in China and India in 20th Century
By Nitin Kumar Bharti and Li Yang:
Abstract. The education system of a country is instrumental in its long-run development. This paper compares the historical evolution of the education systems in the two largest emerging economies- China and India, between 1900 and 2018. We create a novel time-series data of educational statistics related to enrolment, graduates, teachers and expenditure based on historical statistical reports. China adopted a bottom-up approach in expanding its education system, compared to India’s top-down approach in terms of enrolment. While India had a head-start in modern education, it has gradually been overtaken by China- at Primary education in the 1930’s Middle/Secondary level in the 1970s and Higher/Tertiary level in the 2010s. It resulted in the lower cohort-wise average education and higher education inequality in India since 1907. Vocational education is a central component of the Chinese education system, absorbing half of the students in higher education. In India, the majority of the students pursue traditional degree courses (Bachelors, Masters etc.), with 60% in Humanities courses. Though India is known as the “land of engineers”, China produces a higher share of engineers. We conjecture that the type of human capital in China through engineering and vocational education helped develop its manufacturing sector. Utilizing micro-survey data since the 1980s, we show that education expansion has been an inequality enhancer in India. This is due to both the unequal distribution of educational attainment and higher individual returns to education in India.
Interesting throughout, via Pseudoerasmus.
Higher education is getting cheaper
That is the topic of my latest Bloomberg column, here is one excerpt:
There are a lot of numbers, but here is the comparison I find most impressive: Adjusting for grants, rather than taking sticker prices at face value, the inflation-adjusted tuition cost for an in-state freshman at a four-year public university is $2,480 for this school year. That is a 40% decline from a decade ago…
As might be expected, the trajectory for student debt is down as well. About half of last year’s graduates had no student debt. In 2013, only 40% did. That famous saying from economics — if something cannot go on forever, it will stop — is basically true. Due to changes in the formula, aid for Pell Grants is up, which helps to limit both student debt and the expenses of college.
Is quality going down? Probably a bit, but with a caveat:
,,,various adjustments kick in to limit the scope of the potential damage. Rather than cutting classes in computer science, a university might decide (as mine did) not to field a football team. Or a school might rely less on full-time professors and more on adjuncts. That is often a negative, but again schools can and do adjust, for instance by paying their adjuncts more and putting more effort into finding and keeping the good ones. A school might also reduce courses that attract few students and put more emphasis on subject areas with high enrollments.
Granted, none of this is ideal. But such adjustments can keep much of the damage at manageable levels. Many schools also are easing off their DEI bureaucracies.
And students will make adjustments of their own. If their classes give them less than what they want, they may turn more to the internet — to online education or, these days, AI. To argue that a large-language model is not as good as a professor is to miss the point. These innovations only have to make up some of the marginal deteriorations of quality.
With apologies to Peter Thiel, I believe U.S. higher education is going to muddle through.
Where they are headed
The Australian government has pledged to legislate an age limit of 16 years for social media access, with penalties for online platforms that do not comply.
But the Labor government has not spelled out how it expects Facebook, Instagram, TikTok and others to actually enforce that age limit. Anthony Albanese is facing pressure from the Coalition opposition to rush the bill through parliament in the next three weeks, although a federal trial into age assurance technology has not yet commenced.
Albanese and the communications minister, Michelle Rowland, did not rule out the potential for social media users to have their faces subject to biometric scanning, for online platforms to verify users’ ages using a government database, or for all social media users – regardless of age – being subject to age checks, only saying it would be up to tech companies to set their own processes.
Here is the full story. Keep in mind this move, if applied consistently, would eliminate anonymous postings. It also would have to be enforced across a very large number of apps, even for Meta alone. Should everyone’s biometrics be put into what might be China-hackable form? And it means the government — not the parents — is deciding the proper level of social media access for children.
Are the major social media critics for this? Against it? Or are they not so keen to say, one way or the other?
Friday assorted links
The economics of U.S. LNG exports
I assess the climate impact of granting federal approval to all proposed U.S. liquified natural gas (LNG) export terminal projects, which would double U.S. export capacity by 2030. Results indicate a net decrease in global emissions through 2070, primarily due to higher local gas prices in the U.S., leading to lower domestic gas generation and accelerated renewable adoption.
That is from the job market paper of Constanza Abuin, from Harvard University.
Do you want a Democratic or Republican doctor?
Political polarization is increasingly affecting policymaking, but how is it influencing professional decision-making? This paper studies the differences in medical practice between Republican and Democratic physicians over 1999-2019. It links physicians in the Medicare claims data with their campaign contributions to determine their partyalignment. In 1999, there were no partisan differences in medical expenditure perpatient. By 2019, Republican physicians are now spending 13% more, or $70 annually per patient. We analyze four potential sources of this partisan difference: practice characteristics (i.e., specialization and location), patient composition, preferences for financial gain, and beliefs about appropriate care. Even among physicians in the same specialty and location treating patients for the same condition, Republican physicians spend 6% more, especially on elective procedures. Using a movers design, we also find large partisan differences for treating the same patient. We find no evidence that these partisan differences are driven by profit incentives. Instead, the evidence points to diverging beliefs. Republican physicians adhere less to clinical guidelines, consistent with their reported beliefs in prior surveys. The timing of the divergence matches the politicization of evidence-based medicine in Congress. These results suggest that political polarization may lead to partisan differences in the beliefs and behavior of practitioners.
That is from the job market paper of Woojin Kim from UC Berkeley. I found this one of the most interesting job market papers of this year.
The 92nd St. Y
Income Inequality
Robert Kuttner and Tyler Cowen in conversation, moderated by Jeff Greenfield
Thursday assorted links
Prediction Markets for the Win
The prediction markets predicted the election outcome more accurately and more quickly than polls or other forecasting methods, just as expected from decades of research. In this election, however, many people discounted the prediction markets because of large trades on Polymarket. Paul Krugman, for example, wrote:
Never mind the prediction markets, which are thin and easily manipulated.
None of that was true but perhaps that was par for the course. Even some prediction market experts, however, began to wobble under the influence of “whale” manipulation theories. But this story was always shaky. What was the supposed logic?
Few directly articulated the theory—perhaps because it sounds absurd when spelled out. The idea seems to be that whales shifted market odds from 50:50 to 40:60, hoping this would drive more people to vote for Trump. Really? Were voters in Pennsylvania watching Polymarket to decide who to vote for? In a decision market, manipulation might be desirable to a whale (albeit unlikely to succeed), but in prediction markets, this scenario seems dubious: a) people would need to know about these markets, b) they’d need to care about probability shifts on these markets (as opposed to voting say the way their family and neighbors were voting), and c) this would have to be an effective way to spend money to influence votes compared to the myriad other ways of influencing voting. Each step seems dubious.
Alternatively, maybe whales were simply wasting money, “memeing” away millions of dollars? Is that something that whales do? The memeing theory is more plausible with many small traders, not a few whales. Or maybe the whales aimed to spark excitement among the minnows, hoping to build momentum before cashing out. However, exciting small traders to inflate prices and then exiting is risky; the same power that whales have to drive up prices can drive prices down just as quickly, making a profitable exit challenging. In short, while not impossible, the idea of whale-driven manipulation in prediction markets was far-fetched.
In fact, we now know that the biggest whale was moving the markets towards accuracy (against his own interest by the way). In an excellent WSJ article we learn:
The mystery trader known as the “Trump whale” is set to reap almost $50 million in profit after running the table on a series of bold bets tied to the presidential election.
Not only did he see Donald Trump winning the presidency, he wagered that Trump would win the popular vote—an outcome that many political observers saw as unlikely. “Théo,” as the trader called himself, also bet that Trump would win the “blue wall” swing states of Pennsylvania, Michigan and Wisconsin.
Now, Théo is set for a huge payday. He made his wagers on Polymarket, a crypto-based betting platform, using four anonymous accounts. Although he has declined to share his identity, he has been communicating with a Wall Street Journal reporter since an article on Oct. 18 drew attention to his bets.
In dozens of emails, Théo said his wager was essentially a bet against the accuracy of polling data. Describing himself as a wealthy Frenchman who had previously worked as a trader for several banks, he told the Journal that he began applying his mathematical know-how to analyze U.S. polls over the summer.
Here’s the most remarkable bit. Theo commissioned his own polls using a different methodology!
Polls failed to account for the “shy Trump voter effect,” Théo said. Either Trump backers were reluctant to tell pollsters that they supported the former president, or they didn’t want to participate in polls, Théo wrote.
To solve this problem, Théo argued that pollsters should use what are known as neighbor polls that ask respondents which candidates they expect their neighbors to support. The idea is that people might not want to reveal their own preferences, but will indirectly reveal them when asked to guess who their neighbors plan to vote for.
…In an email, he told the Journal that he had commissioned his own surveys to measure the neighbor effect, using a major pollster whom he declined to name. The results, he wrote, “were mind blowing to the favor of Trump!”
Théo declined to share those surveys, saying his agreement with the pollster required him to keep the results private. But he argued that U.S. pollsters should use the neighbor method in future surveys to avoid another embarrassing miss.
Thus, a big win for prediction markets, for Polymarket and for GMU’s Robin Hanson, the father of prediction markets, whose work directly influenced the creation of Polymarket.