Category: Science
A Fine Theorem on Parag Pathak, the recent Clark winner
This award strikes me as the last remaining award, at least in the near term, from the matching/market design boom of the past 20 years. As Becker took economics out of pure market transactions and into a wider world of rational choice under constraints, the work of Al Roth and his descendants, including Parag Pathak, has greatly expanded our ability to take advantage of choice and local knowledge in situations like education and health where, for many reasons, we do not use the price mechanism. That said, there remains quite a bit to do on understanding how to get the benefits of decentralization without price – I am deeply interested in this question when it comes to innovation policy – and I don’t doubt that two decades from now, continued inquiry along these lines will have fruitfully exploited the methods and careful technique that Parag Pathak embodies.
The post is excellent throughout.
A history of randomized assignment in the social sciences
Although the concept of randomized assignment to control for extraneous factors reaches back hundreds of years, the first empirical use appears to have been in an 1835 trial of homeopathic medicine. Throughout the 19th century, there was primarily a growing awareness of the need for careful comparison groups, albeit often without the realization that randomization could be a particularly clean method to achieve that goal. In the second and more crucial phase of this history, four separate but related disciplines introduced randomized control trials within a few years of one another in the 1920s: agricultural science, clinical medicine, educational psychology, and social policy (specifically political science). Randomized control trials brought more rigor to fields that were in the process of expanding their purviews and focusing more on causal relationships. In the third phase, the 1950s through the 1970s saw a surge of interest in more applied randomized experiments in economics and elsewhere, in the lab and especially in the field.
That is from a Julian C. Jamison paper done at the World Bank, via various people in my Twitter feed.
Models as indexing, and the value of Google
There are many arguments for the use of models in economics, including notions of rigor and transparency, or that models can help you to see relationships you otherwise might not have expected. I don’t wish to gainsay those, but I thought of another argument yesterday. Models are a way of indexing your thoughts. A model can tell you which are the core features of your argument and force you to give them names. You then can use those names to find what others have written about your topic and your mechanisms. In essence, you are expanding the division of labor in science more effectively by using models.
This mechanism of course requires that models are a more efficient means of indexing thoughts than pure words or propositions alone. In this view, it is often topic names or book indexes or card catalogs that models are competing with, not verbal economics per se.
The existence of Google therefore may have lowered the relative return to models. First, Google searches by words best of all. Second and relatedly, if you have written only words Google will help you find the related work you need, scholar.google.com kicks in too. In essence, there is a new and very powerful way of finding related ideas, and you need not rely on the communities that get built around particular models (though those communities largely will continue).
It is notable that open access, on-line economics writing doesn’t use models very much and is mostly content to rely on words and propositions. There are several reasons for this, but this productivity shock to differing methods of indexing may be one factor.
Still, it is not always easy to search by words. Many phrases — consider say “free will” — do not through search engines discriminate very well on the basis of IQ or rigor.
Indian population bottlenecks
…we found that West Eurasian-related mixture in India ranges from as low as 20 percent to as high as 80 percent…
Groups of traditionally higher social status in the Indian caste system typically have a higher proportion of ANI [Ancestral North Indians] ancestry than those of traditionally lower social status, even within the same state of India where everyone speaks the same language. For example, Brahmins, the priestly caste, tend to have more ANI ancestry than the groups they live among, even those speaking the same language.
It also seems that a disproportionate share of the ANI genetic input came from males. Furthermore:
Around a third of Indian groups experienced population bottlenecks as strong or stronger than the ones that occurred among Finns or Ashkenazi Jews.
Many of the population bottlenecks in India were also exceedingly old. One of the most striking we discovered was in the Vysya of the souther Indian state of Andhra Pradesh, a middle caste group of approximately five million people whose population bottleneck we could date…to betweenthree thousand and two thousand years ago.
The observation of such a strong population bottleneck among the ancestors of the Vysya was shocking. It meant after the population bottleneck, the ancestors of the Vysya had maintained strict endogamy, allowing essentially no genetic mixing into their group for thousands of years.
And the Vysya were not unique. A third of the groups we analyzed gave similar signals, implying thousands of groups in India like this…long-term endogamy as embodied in India today in the institution of caste has been overwhelming important for millennia.
…The truth is that India is composed of a large number of small populations.
That is all from David Reich’s superb Who We Are and How We Got Here: Ancient DNA and the New Science of the Human Past. Here is my earlier post on the book.
The importance of local milieus
Using data on the entire population in combination with data on almost all individuals in Sweden listed as inventors, we study how the probability of being listed on a patent as inventor is influenced by the density of other future inventors residing in the same region. In this process, we control for demographic and sector effects along with the educational characteristics of parents. This approach allows us to trace how location history influences individuals’ inventive capacity. We focus on three types of influences: (a) future inventors in the municipality around the time of birth, (b) future inventors around the time of graduation from high school and (c) future inventors at graduation from higher education. We find suggestive evidence that co-locating with future inventors may impact the probability of becoming an inventor. The most consistent effect is found for place of higher education; some positive effects are also evident from birthplace, whereas no consistent positive effect can be derived from individuals’ high school location. Therefore, the formative influences mainly deriving from family upbringing, birth region and from local milieu effects arising from a conscious choice to attend a higher education affect the choice of becoming an inventor.
Here is the article, “How important are local inventive milieus: The role of birthplace, high school and university education,” by Olof Ejermo and Høgni Kalsø Hansen, via Ben Southwood.
Software engineer and psychologist wanted
AI Grant (aigrant.org) is a distributed AI research lab. Their goal is to find and fund the modern day Einsteins: brilliant minds from untraditional backgrounds working on AI.
They need a software engineer. The qualifications are twofold: an intrinsic interest in the problem of identifying talented people across the world, and a demonstrated ability to ship software projects without much supervision. This doesn’t have to be through traditional means. It could just be side-projects on Github.
They’re also looking for a psychologist with experience in personality and IQ modeling.
Bay Area location is a big plus, but not a requirement. If you’re interested in learning more, email team@aigrant.org with information about yourself.
This is not a paid ad, but I am seeking to do a favor for the excellent and highly talented Daniel Gross (and perhaps for you), with whom you would get to work. Do please mention MR if you decide to apply!
Cross-cultural digital instruction
Comparative ethnographic analysis of three middle schools that vary by student class and race reveals that students’ similar digital skills are differently transformed by teachers into cultural capital for achievement. Teachers effectively discipline students’ digital play but in different ways. At a school serving working-class Latino youth, students are told their digital expressions are irrelevant to learning; at a school with mostly middle-class Asian American youth, students’ digital expressions are seen as threats to their ability to succeed academically; and at a private school with mainly wealthy white youth, students’ digital skills are positioned as essential to school success. Insofar as digital competency represents a kind of cultural capital, the minority and working-class students also have that capital. But theirs is not translated into teacher-supported opportunities for achievement.
Here is the AJS piece, by Matthew H. Rafalow. For the pointer I thank Kevin Lewis.
Do earthquakes make men more risk tolerant?
We investigate whether individuals’ risk preferences change after experiencing a natural disaster, specifically, the 2011 Great East Japan Earthquake. Exploiting the panels of nationally representative surveys on risk preferences, we find that men who experienced greater intensity of the earthquake became more risk tolerant a year after the Earthquake. Interestingly, the effects on men’s risk preferences are persistent even five years after the Earthquake at almost the same magnitude as those shortly after the Earthquake. Furthermore, these men gamble more, which is consistent with the direction of changes in risk preferences. We find no such pattern for women.
That is from a newly published paper by Chie Hanaoka, Hitoshi Shigeoka, and Yasutora Watanabe. What else will have this effect?
*Who We Are and How We Got Here*, by David Reich
This is a truly excellent work, readable and informative at A to A+ quality, and the subtitle is Ancient DNA and the new Science of the Human Past. It has occasioned some public controversy for its discussion of race and genetics, but most of all this is a book about how science is done. For instance, the page and a half discussion of how researchers try to ensure that human DNA does not contaminate Neanderthal DNA is just beautiful.
Here is one good summary passage:
The case of the Ancient North Eurasians showed that while a tree is a good analogy for the relationships among species — because species rarely interbreed and so like real tree limbs are not expected to grow back together after they branch — it is a dangerous analogy for human populations. The genome revolution has taught us that great mixtures of highly divergent populations have occurred repeatedly. Instead of a tree, a better metaphor may be a trellis, branching and remixing far back into the past.
Here is another excerpt of note:
Analyzing our data, he [Iosif Lazaridis] found that about ten thousand years ago there were at least four major populations in West Eurasia — the farmers of the Fertile Crescent, the farmers of Iran, the hunter-gatherers of central and western Europe, and the hunter-gatherers of eastern Europe. All these populations differed from one another as much as Europeans differ from East Asians today.
The concept of “ghost populations” will enter your mental conceptual vocabulary. And:
The extraordinary fact that emerges from ancient DNA is that just five thousand years ago, the people who are now the primary ancestors of all extant northern Europeans had not yet arrived.
Most of all, this is a science book, not a “race book.” (“Having been immersed in the ancient DNA revolution for the past 10 years, I am confident that anyone who pays attention to what it is finding cannot come away feeling affirmed in racist beliefs.”) You may know that Reich is a Professor of Genetics at Harvard Medical School.
Here is his earlier NYT essay (though I think the very first link in this post is the best place to start, do read that carefully), well done but not quite representative of the book either. You can buy it here, this is definitely one of the books of the year and one of the best popular science books of any year.
*Cognitive Gadgets*
The author is Cecilia Heyes, and the subtitle is The Cultural Evolution of Thinking, published by Harvard/Belknap. It is not always a transparent read, but this is an important book and likely the most thoughtful of the year in the social sciences.
From the book’s home page:
…adult humans have impressive pieces of cognitive equipment. In her framing, however, these cognitive gadgets are not instincts programmed in the genes but are constructed in the course of childhood through social interaction. Cognitive gadgets are products of cultural evolution, rather than genetic evolution. At birth, the minds of human babies are only subtly different from the minds of newborn chimpanzees. We are friendlier, our attention is drawn to different things, and we have a capacity to learn and remember that outstrips the abilities of newborn chimpanzees. Yet when these subtle differences are exposed to culture-soaked human environments, they have enormous effects. They enable us to upload distinctively human ways of thinking from the social world around us.
The key substantive points from this are malleability and speed of evolution, and overall in her theory there is a much lower reliance on cognitive instincts and thus a fundamentally different account of social evolution: “In contrast, the cognitive gadgets theory applies cultural evolutionary theory to the mechanisms of thought — the mental processes that generate and control behavior.”
And “…social interaction in infancy and childhood produces new cognitive mechanisms; it changes the way we think.”
The chapter on imitation is the best appendage to Girard on memesis I know. One interesting point is that most people find it quite hard to imitate how they look to others when say they tell a joke or make love. To imitate successfully, you need to develop particular sensorimotor capacities. Otherwise, you can be thwarted by a kind of “correspondence” problem, not knowing how the objective and subjective experiences of imitation match up properly. This too we learn through cultural gadgets.
Mindreading is also a mental gadget, it must be learned, and it is surprisingly similar to print reading. In an odd twist on Julian Jaynes, Heyes suggests that humans five or six thousand years ago may not have had this capacity very strongly. And as with print reading, there is cross-cultural diversity in mindreading. There is no mindreading instinct and we all must learn it, autistics too.
What about language? Rather than Chomsky’s Universal Grammar, there are instead “domain-general processes of sequence learning.” This in turn leads to a complex and quite interesting take on how, while non-human animals do also have language, it is quite different from ours (p.187).
Most generally, if someone is trying to explain X, maybe both genetic/instinct and cultural evolution accounts of X are wrong — try a cultural gadget approach! And think of this book as perhaps the best attempt so far to explain the weirdness of humans, relative to other animals.
Note also that in this view, humanity is relatively vulnerable to cultural catastrophes, as we cannot simply bounce back using enduring instincts. Furthermore, social media may indeed matter a great deal, and in revisionist terms some parts of Marx are not as crazy as they may seem (my point this latter one, not hers).
I need more time (years?) to digest the contents of this book, and decide how much I agree. It is somehow neither hard nor easy reading, but most MR readers should be able to make their way through it. Highly recommended, it is likely to prove one of the most thought-provoking books of the year.
Buy it on Amazon here. Here is a Heyes lecture on related ideas, also click through to part II.
My Conversation with Martina Navratilova
Here is the podcast and transcript, Martina was in top form and dare I say quick on her feet? Here is part of the summary:
In their conversation, she and Tyler cover her illustrious tennis career, her experience defecting from Czechoslovakia and later becoming a dual citizen, the wage gap in tennis competition and commentary, gender stereotypes in sports, her work regimen and training schedule, technological progress in tennis, her need for speed, journaling and constant self-improvement, some of her most shocking realizations about American life, the best way to see East Africa, her struggle to get her children to put the dishes in the dishwasher, and more.
Here is one bit:
NAVRATILOVA: I just wanted to leave no stone unturned, really. The coach, obviously, was technique and tactics. The physical part was training, working very hard. I’ll give you my typical day in a minute. The eating was so that I could train hard and not get injured. So it all came together.
The typical day, then, when I really was humming was four hours of tennis, 10:00 to 2:00, two hours of drills and maybe two hours of sets. Then I would do some running drills on the court for 15, 20 minutes, sprints that if I did them now, I wouldn’t be able to walk the next day.
[laughter]
NAVRATILOVA: You know, 15- to 30-second sprinting drills. Then we would eat lunch. Then I would go either play basketball full-court, two on two for an hour and a half or little man-big man. It’s two on one. I don’t know, those people that play basketball. You just run. You just run.
COWEN: Which one were you?
NAVRATILOVA: It switches. Whoever has the ball is the little man. No, whoever has the ball, it’s one against two. Then you play little man, the person plays defense, and then the big man plays center. It’s not two on one, it’s one against one and then one. Then whoever gets the ball goes the other way. It’s run, run, run.
Then I would lift weights and have dinner either before lifting weights or after. So it was a full day of training.
COWEN: What about 9:00 A.M. to 10:00 A.M.?
And this:
COWEN:Billie Jean King once suggested that you use writing in a journal every day to help you accomplish your goals. How does that work for you? What is it you do? Why do you think it works?
NAVRATILOVA: It worked because it really centers you. It narrows it down, whatever long-term goal you have. It becomes more real and more current because it narrows it down in that, “What do you need to do today?” and “Did you accomplish that goal?” You have a big goal. You break it into smaller goals, into smaller goals, until you get into, “OK, what do I do today to get to that goal?”
…Try to be honest with yourself. Be honest but also be nice to yourself. You see that with most champions. They’re perfectionists. You beat yourself up too much. I preach and I try to strive for excellence rather than perfection.
If you strive for excellence, perfection may happen. [laughs] It’s good enough to be excellent. That’s good enough. You don’t need to be perfect because perfection just happens by accident.
I asked her this:
COWEN: What was it like to go skiing with Donald Trump?
My favorite part was this:
NAVRATILOVA: Tyler, you need to drink more water. You’re not hydrating at all.
Remember, above all else, sports is cognitive! These are some of the smartest humans of our time, even if it is not always the kind of intelligence you respect most.
What should I ask Balaji Srinivasan?
I will be doing a Conversation with him, no associated public event. Here is his home page, here is his bio:
Balaji S. Srinivasan is the CEO of Earn.com and a Board Partner at Andreessen Horowitz. Prior to taking the CEO role at Earn.com, Dr. Srinivasan was a General Partner at Andreessen Horowitz. Before joining a16z, he was the cofounder and CTO of Founders Fund-backed Counsyl, where he won the Wall Street Journal Innovation Award for Medicine and was named to the MIT TR35.
Dr. Srinivasan holds a BS, MS, and PhD in Electrical Engineering and an MS in Chemical Engineering from Stanford University. He also teaches the occasional class at Stanford, including an online MOOC in 2013 which reached 250,000+ students worldwide.
His latest Medium essay was on ICOs and tokens. I thank you all in advance for your wise counsel.
Prisoners who study prison
After leaving prison, some ex-convicts are becoming academics themselves. There is a growing convict criminology group, which has members in countries around the world.
…Stephen Richards spent nine years in prison for conspiracy to distribute marijuana, and is now professor emeritus of criminology at the University of Wisconsin at Oshkosh.
He was arrested while he was a college student, and finished his degree by correspondence while in U.S. federal prison. After he was released from prison, he went directly to graduate school and completed a Master’s degree and a PhD.
“Five years out of prison, I was a professor, and I became one of the first convict criminology professors,” he says.
Richards’s experiences in jail made him want to work to fix the system once he got out.
“Part of being a convict criminologist is realizing that you know something that most academics in the social sciences don’t know. You’ve got inside information about what’s wrong with the criminal justice system — literally, inside. You know what a failure the system is, and you want to do something about it,” he says.
That is from CBC Radio, via Michelle Dawson. And here is the Journal of Prisoners on Prisons.
Longevity FAQ
From Laura Deming, you will find it here, essential reading for our time. Here is one bit:
Senescence
at a glance: a fraction of your cells get older than the others, so we’d like to eliminate them
As you get old, so do your cells. But some of your cells get old in a way that is much worse than the others. You may have heard of a thing called telomerase. If you remember correctly, it’s the thing that keeps the end of your DNA long enough that your cells can still divide. When one of your cells runs out of telomerase, it can’t make many more copies of itself. If the cell sticks around, refuses to die even when it stops working, and starts secreting signals to the immune system, we call that a ‘senescent cell‘.
What happens when you get rid of these cells? Some animals that age faster than normal have a lot of these ‘senescent cells’ and are good experimental models in which to ask that question. In 2011, a group from the Mayo Clinic cleared out many of the senescent cells in one of those animal models, and found that the resulting mice were healthier in old age (among other things, they did not get cataracts and bent spines, which typically emerge in old age). In 2016, the same investigators found that getting rid of senescent cells in normal mice made them live a longer healthy lifespan. Knocking out senescent cells is tricky, because they don’t have many unique identifiers. Companies are working to either find things empirically that kill senescent cells, or figure out specific mechanisms by which to try to destroy them.
It starts off like this:
Hi! I’m Laura Deming, and I run Longevity Fund. I spend a lot of time thinking about what could increase healthy human lifespan. This is my overview of the field for beginners.
Does it end with you living to 129? I genuinely do not know.
The World’s Biggest Field Experiment
A new paper (another summary) in Nature reports on what is perhaps the world’s biggest field experiment which has successfully shown how to, at scale, increase crop yields and reduce fertilizer usage in China. The scope of the 10 year experiment is astounding. The researchers first conducted thousands of field experiments all over China to discover and validate best practices:
A total of 13,123 site years of field trials were conducted from 2005 to
2015 for the three crops (n=6,089 for maize, 3,300 for rice and 3,734 for wheat), with sites spread across all agro-ecological zones…Each field trial included two types of management: conventional farmers’ practice (control) and ISSM-based recommendations (treatment; developed specifically for a given area). The recommended practices were discussed with local experts and participating farmers. Adjustments were made when necessary. Finally, the agreed-upon management technologies were implemented in the fields by the farmer; the collaborators provided guidance on-site during key operations, such
as sowing, fertilization, irrigation and harvest. Campaign collaborators recorded fertilizer rate, pesticide and energy use, and calculated nutrient application rate. At maturity, grain yield and above ground biomass were sampled by the collaborators for plots with a size of 6m^2 for wheat and rice, and 10m^2 for maize. Plant samples were dried at 70 °C in a forced-draft oven to constant weight, and grain yield was standardized at 14% moisture for all crops.
With validated best practices in hand the researchers and tens of thousands of collaborators then fanned out across the country to convince farmers to adopt the best practices.
During the campaign, about 14,000 training workshops, 21,000 field days, and more than 6,000 site demonstrations were organized by campaign staff; more than 337,000 pamphlets were distributed….During the campaign, we also encountered barriers and experienced challenges. For example, we observed that some farmers appeared indifferent during some outreach events. We later learned that it was mainly, because they could not comprehend the scientific content that we were trying to deliver. We solved the problem by having local (county or township) agents acting as an on-site ‘interpreter’ in ways that speaks/connects with those farmers.
This was amusing:
It is also worth noting that the interests of agribusinesses do not always align with those of our campaign staff. For example, one of our main strategies used in the campaign was to select a site (for example, a village) for a given area, establish the base with field demonstrations of ISSM-based practices, then attract and engage more farmers from the same as well as neighbouring villages, creating a snowballing and lasting effect. But sometimes, our partners in the private sector were more interested in changing sites so as to reach more farmer-clients. Vigorous debates and discussion ensued. Eventually, the private sector personnel conformed to our reasoned schemes while using the established sites as demonstrations for visitors from other areas.
Outputs and inputs among the treatment and control farmers were then measured (here I would have liked more information about the randomization. A lot can go wrong or be mismeasured at this stage.).
Farmers conducted all field operations. Campaign collaborators and/or extension agents were responsible for information and data collection. Typically, 10–30 farmers were randomly selected per ISSM-adopting site; another group of randomly selected 10–30 farmers from a nearby village without ISSM intervention served as a control/comparison. From the selected pool of farmers (roughly 14,600 paired data points), information on key management practices were obtained through a questionnaire survey, including crop varieties, planting densities, planting dates, fertilizer rates and harvest dates. For some sites, grain yields were directly measured in the same way as the field trials (see ‘Field trials’) for the selected 10–30 farmers. Yield and nitrogen rate were then averaged for each site.
The results were impressive.
Aggregated 10-year data showed an overall yield improvement of 10.8–11.5% and a reduction in the use of nitrogen fertilizers of 14.7–18.1%, when comparing ISSM-based interventions and the prevailing practices of the farmers. This led to a net increase of 33 Mt grains and a decrease of 1.2 Mt nitrogen fertilizer use during the 10-year period, equivalent to US$12.2 billion.
The entire experiment cost on the order of $56 million and generating $12.2 billion dollars of increased output, not including any environmental gains.
As if this weren’t enough the researchers then surveyed over 8 million smallholder farmers in China to estimate how much output could increase if the intervention were fully scaled.
What’s especially encouraging about this project is that no new technologies, seeds or infrastructure was involved–just basic science and a tremendous outreach campaign. Moreover, since the campaign increased profits it may continue to generate gains in the future even without further intervention as the practices spread. Repeated interventions will be necessary as climate changes, however. Information technology may makes this easier. China can be intimidating.