Month: October 2022

The wisdom of Daniel Gross

by  Tyler Cowen

Now in 2017, a bunch of people, each of which now has their own company, the new PayPal Mafia is the Transformer Mafia, wrote this paper called Attention is All You Need, which at the time was mostly ignored by the rest of the world, and they came up with a way to effectively parallelize this training, and enable us to create models that are much larger, and as a byproduct are able to store more context tokens over and over, but effectively more words and effectively be able to predict more words to you.

The paper was mostly ignored when it came out — I thought it was neat, I don’t know that I made much of it. Google at the time had developed this pretty large model based on the paper that it didn’t release for various reasons we can touch on. Then OpenAI really productized that paper with GPT-2 and 3, general purpose transformer, that transformer is from that paper from Attention is All You Need. They were able to build these successively larger and larger models because they were able to parallelize training. These models now, GPT-3, is considered state-of-the-art, although I think our grandchildren will look at that the same way as we look at tube television.

That is from the new Ben Thompson interview with Daniel and Nat Friedman, and yes I do subscribe to Ben and pay for it.

Is there a neglect of low-probability gains?

by  Tyler Cowen

Seven preregistered studies (N = 2,890, adult participants) conducted in the field, in the lab, and online documented opportunity neglect: a tendency to reject opportunities with low probability of success even when they come with little or no objective cost (e.g., time, money, reputation). Participants rejected a low-probability opportunity in an everyday context (Study 1). Participants also rejected incentive-compatible gambles with positive expected value—for both goods (Study 2) and money (Studies 3–7)—even with no possibility of monetary loss and nontrivial rewards (e.g., a 1% chance at $99). Participants rejected low-probability opportunities more frequently than high-probability opportunities with equal expected value (Study 3). Although taking some real-life opportunities comes with costs, we show that people are even willing to incur costs to opt out of low-probability opportunities (Study 4). Opportunity neglect can be mitigated by highlighting that rejecting an opportunity is equivalent to choosing a zero probability of success (Studies 6–7).

That is from new research by Emily Prinsloo, Kate Barasz, and Michael I. Norton, via the excellent Kevin Lewis.

My Conversation with the excellent Walter Russell Mead

by  Tyler Cowen in

Here is the audio and transcript, here is the summary:

He joined Tyler to discuss how the decline of American religiosity has influenced US foreign policy, which American presidents best and least understood the Middle East, the shrewd reasons Stalin supported Israel, the Saudi secret to political stability, the fate of Pakistan, the most likely scenario for China moving on Taiwan, the gun pointed at the head of German business, the US’s “murderous fetishization of ideology over reality” in Sub-Saharan Africa, the inherent weakness in having a foreign policy establishment dominated by academics, what he learned from attending the Groton School, and much more.

Here is one excerpt:

COWEN: How would you change or improve the training that goes into America’s foreign policy elite?

MEAD: Well, I would start by trying to draw people’s attention to that, over the last 40 years, there’s been an enormous increase in the number of PhD grads engaged in the formation of American foreign policy. There’s also been an extraordinary decline in the effectiveness of American foreign policy. We really ought to take that to heart.

COWEN: Do you think of it as an advantage that you don’t have a PhD?

MEAD: Huge advantage.

COWEN: How would you describe that advantage?

MEAD: I don’t really believe in disciplines. I see connections between things. I start from reality. I’m not trying to be anti-intellectual here. You need ideas to help you organize your perceptions of reality. But I think there’s a tendency in a lot of social science disciplines — you start from a bunch of really smart, engaged people who have been thinking about a set of questions and say, “We’ll do a lot better if we stop randomly thinking about everything that pops up and try, in some systematic way, to organize our thinking of this.”

I think you do get some gains from that, but you see, over time, the focus of the discipline has this tendency to shift. The discipline tends to become more inward navel-gazing. “What’s the history of our efforts to systematize our thinking about this?” The discipline becomes more and more, in a sense, ideological and internally focused and less pragmatic.

I think that some of the problem, though, is not so much in the intellectual weaknesses of a lot of conventional postgrad education, but simply almost the crime against humanity of having whole generations of smart people spend the first 30, 35 years of their lives in a total bubble, where they’re in this academic setting, and the rule . . . They become socialized into the academy, just as much as prisoners get socialized into the routines of a prison.

The American academy is actually a terrible place for coming to understand how world politics works. Recently, I had a conversation with an American official who was very proud of the way that the US had broken the mold by revealing intelligence about Russia’s plans to invade Ukraine, and pointed out how that had really helped build the NATO coalition against Russian aggression, and so on.

So far as he goes, it’s true. But I said, however, if you really look at the total message the US was projecting to Russia in those critical months, there were two messages. One is, “We’ve got great intelligence on you. We actually understand you much better than you think.” It was shocking. I think it shocked the Russians. But on the other hand, we’re saying, “We think you’re going to win quickly in Ukraine. We’re offering Zelenskyy a plane ride out of Kyiv. We’re pulling out all our diplomats and urging other countries to pull out their diplomats.”

The message, actually the totality of the message that we sent to Putin is, “You are going to win if you do this.”

And this, on what makes for talent in the foreign policy arena:

…you can’t know too much history. A hunger for travel. I think too many foreign policy types don’t actually get out into the field nearly as much as they should. Curiosity about other cultures. A strong grounding in a faith of your own, which can be a secular ideology, perhaps, in some cases, but more often is likely to be a great religious tradition of some kind.

A very good conversation.  And I am happy to recommend Walter’s new book The Arc of a Covenant: The United States, Israel, and the Fate of the Jewish People.

India: The Revolution in Private Schooling

by  Alex Tabarrok in

A whopping 50%+ of secondary school students in India are educated in private schools. Do private schools increase human capital or merely skim the best students? My paper, Private Education in India: A Novel Test of Cream Skimming made a simple but telling point:

…As the private share of school enrollment increases simple cream skimming becomes less plausible as the explanation for a higher rate of achievement in private schools. If the private schools cream skim when they are at 10% of public school enrollment how much cream can be left in the public school pool when the private schools account for 60% of total enrollment? Thus, if this simple form of cream skimming is the explanation for the higher achievement rate in private schools, we would expect the “private effect,” the difference between private and public scores, to be smaller in regions with a high
share of private schooling.

In fact, what I find is the private advantage, although larger in districts with smaller shares of private schooling (suggesting some skimming), stabilizes and doesn’t disappear even as the share of private schooling heads towards 100%. I also show that mean scores across all students, public and private, increase with the share of private schooling which is inconsistent with cream skimming (which predicts a constant mean). At right a picture showing that private scores continue to outpace public scores even in districts where private schools educate a majority or larger share of students.

In a new paper, Bagde, Epple and Taylor study 4 million students in thousands of villages in India during 2004-2014. In the early years of the study, none of the villages have private schools but entry starts to occur in 2007-2009 and the authors look at who switches to private schools. They find significant selection from higher income, higher caste, higher ability, and males towards private schools but no evidence that public school students are harmed.

The authors give a nod to the possibility that stratification could generate problems down the line if it increases inequality but they don’t mention the key point that, as with arguments for cream skimming, stratification concerns diminish the more students are in private schools and disappear altogether if 100% of students are in private schools.

More generally, India is pioneering private education on a grand scale and the entire world should pay attention to these innovations.

Addendum: See also my previous post on a key paper by Muralidharan and Sundararaman, Private Schooling In India: Results from a Randomized Trial.

“But are you long volatility?”

by  Tyler Cowen in

As many of you know, when I confront mega-pessimists I like to ask them “But are you short the market?”  Not once in my life have I heard a satisfactory rejoinder to this query.  (The last answer I heard was “I will be!”.)

I now have a new question for those who see a reasonably high likelihood of AGI.  Of course AGI could wreck the world, make the world a whole lot better, or simply overturn multiple sectors of the economy, in both good and bad ways.  If you believe in AGI, typically you believe it will matter a lot, though there is considerable disagreement on the cost-benefit ratio.

So basically you should go long volatility, both of the aggregate market and sector-by-sector.

Can I call them AGIers?  (“Aggies” is already taken.)  AGIers ought to be long volatility.

Are they?  I have started asking this question, and I plan to continue the practice.

The market for property insurance vs. climate change

by  Tyler Cowen in

That is the topic of my latest Bloomberg column, here is one bit:

One of the classic rejoinders to worries about climate change is the claim that people can move out of highly vulnerable areas into safer areas. Maybe the world will not be willing to accept hundreds of millions of climate-change refugees, but within the US, perhaps people can move from storm-prone Florida to the northern Midwest, or to wherever might prove appropriate, including safer parts of Florida. The US, after all, has a longstanding tradition of individual mobility. And many parts of the country have the space and infrastructure for additional residents.

For such migration to have any effect on the costs of climate change, however, price signals have to be active and relatively undistorted. That is, some set of market prices has to be giving people impetus to leave one place for another. And policymakers have not been letting insurance markets perform their proper work in this regard.

And on the details:

Currently the market for Florida property insurance is in a pretty bad way. This year six relevant insurance companies went insolvent, and for Florida underwriting losses have run more than $1 billion for each of the last two years. Not surprisingly, insurers have been cutting back their coverage in the state or leaving altogether. The end result is that homeowners are finding it much harder to get coverage and finding it much more expensive when they do. None of this should come as a surprise, given the immense damage wrought by Hurricane Ian and previous storms.

Yet politics is stifling market adjustments. Florida has a state-run insurer of last resort, called Citizens Property Insurance Corp. Not surprisingly, that insurer has financial problems of its own, and in May Governor Ron DeSantis oversaw an additional $2 billion in reinsurance support for the company’s efforts. In other words, the state government is stifling the market signals that might induce some of the state’s homeowners to leave for drier pastures.

But don’t put your hopes in the Florida gubernatorial election. DeSantis’s Democratic rival, Charlie Crist, has criticized the governor for not doing more on the property insurance front and has proposed 90-day emergency insurance coverage for residents. That would stifle market incentives all the more.

I should note that water subsidies for the Southwest are another example of the same general phenomenon.

That was then, this is now — Iranian edition

by  Tyler Cowen in

Women protestors were part of the constitutional movement from its earliest stages.  They facilitated the strikes, lent their moral and financial support to the constitutionalists, and defended them physically against the forces of the shah.  In 1905 women reportedly created human barriers and protected the ‘ulama who had taken sanctuary at the Shah ‘Abd al-‘Azim Shrine from the armed government forces.  In the summer of 1906, when nationalists obtained sanctuary at the garden of the British legation, several thousand women gathered to join the strikers.

That is from Janet Afary’s quite interesting The Iranian Constitutional Revolution, 1906-1911, Grassroots Democracy, Social Democracy, & the Origins of Feminism.

Tuesday assorted links

by  Tyler Cowen

1. How to run surveys.

2. German wives listen more to German husbands than vice versa.

3. “Currently, women are 3-15 times more likely to be selected as members of the AAAS and NAS than men with similar publication and citation records.”

4. Haiti is truly collapsing.  Yet few people wish to talk about what happens when a nation-state is no longer a viable nation-state.

5. What is going wrong in Principles classes?

6. Part of why I think AGI will prove difficult, namely that human use their whole body to compute.

7. Summers and Biden.

8. Good review of Vesper, a good movie on the big screen.

Are Immigrants more Left leaning than Natives?

by  Tyler Cowen in

We analyze whether second-generation immigrants have different political preferences relative to observationally identical children of citizens in the host countries. Using data on individual voting behavior in 22 European countries between 2001 and 2017, we characterize each vote on a left-right scale based on the ideological and policy positions of the party receiving the vote. In the first part of the paper, we characterize the size of the “left-wing bias” in the vote of second-generation immigrants after controlling for a large set of individual characteristics and origin and destination country fixed effects. We find a significant left-wing bias of second-generation immigrants, comparable in magnitude to the left-wing bias associated with living in urban (rather than rural) areas. We then show that this left-wing bias is associated with stronger preferences for inequality-reducing government intervention, internationalism and multiculturalism. We do not find that second-generation immigrants are biased towards or away from populist political agendas.

That is from a new paper by Simone Moriconi, Giovanni Peri, and Riccardo Turati.

The polity or is it culture that is Swiss?

by  Tyler Cowen

Swiss consumer price inflation slowed to 3.3% in September but still surpassed the Swiss National Bank’s 0-2% target range for the eighth month in a row.

Economists had expected the year-on-year rate to hold steady at August’s 3.5%, a 29-year high.

The index fell 0.2% versus August given falling prices for fuels, heating oil, hotels and supplementary accommodation. Prices for clothing and footwear increased.

Here is the full story, via Johann C.

Monday assorted links

by  Tyler Cowen

1. The case for education, OLS vs. causal identification techniques.

2. Obamacare and the median voter theorem.

3. Why did the proposed ban on Congressional stock trades fail?

4. Claims about nuclear zugzwang.  And uncertainty about the effects of nuclear explosions.

5. Donne/Rundell fashion show?

6. Proposals for reforming universities (overall I am myself not so convinced by these; it is “ground troops” that matter).