Category: Data Source
Agriculture matters, the past matters
We match individual-level survey data with information on the historical lifeways of ancestors, focusing on Africa, where the transition away from such modes of production began only recently. Within enumeration areas and occupational groups, we find that individuals from ethnicities that derived a larger share of subsistence from agriculture in the pre-colonial era are today more educated and wealthy. A tentative exploration of channels suggests that differences in attitudes and beliefs as well as differential treatment by others, including less political power, may contribute to these divergent outcomes.
That is from a recent paper by Michalopoulos, Putterman, and Weil. Here are video and ungated versions.
Do experts make more errors when they are losing or behind?
Although Stockfish and Komodo have differences in their evaluation scales—happily less pronounced than they were 1 and 2 years ago—they agree that the world’s elite made six times more large errors when on the lower side of equality.
We don’t know how general this phenomenon is, but interestingly it seems to hold much more strongly for top players than for weak players. That is from chess of course.
Here is much more detail from Ken Regan, along with some suggested hypotheses and resolutions.
Mexico (America) fact of the day
Mexican non-oil exports to USA in December (y/y): -4.5%. Excluding autos: -8.7%.
That is from Genevieve Signoret, via this source.
It’s funny how these numbers seem to indicate someone is starting to enter a recession. Who might that be? Maybe it’s just noise, I don’t see any other mediocre economic reports wandering around these parts… Or maybe it’s Mexico that’s the problem…
What will be the impact of machine learning on economics?
The excellent Susan Athey addresses that question on Quora, here is one excerpt:
Machine learning is a broad term; I’m going to use it fairly narrowly here. Within machine learning, there are two branches, supervised and unsupervised machine learning. Supervised machine learning typically entails using a set of “features” or “covariates” (x’s) to predict an outcome (y). There are a variety of ML methods, such as LASSO (see Victor Chernozhukov (MIT) and coauthors who have brought this into economics), random forest, regression trees, support vector machines, etc. One common feature of many ML methods is that they use cross-validation to select model complexity; that is, they repeatedly estimate a model on part of the data and then test it on another part, and they find the “complexity penalty term” that fits the data best in terms of mean-squared error of the prediction (the squared difference between the model prediction and the actual outcome). In much of cross-sectional econometrics, the tradition has been that the researcher specifies one model and then checks “robustness” by looking at 2 or 3 alternatives. I believe that regularization and systematic model selection will become a standard part of empirical practice in economics as we more frequently encounter datasets with many covariates, and also as we see the advantages of being systematic about model selection.
…in general ML prediction models are built on a premise that is fundamentally at odds with a lot of social science work on causal inference. The foundation of supervised ML methods is that model selection (cross-validation) is carried out to optimize goodness of fit on a test sample. A model is good if and only if it predicts well. Yet, a cornerstone of introductory econometrics is that prediction is not causal inference, and indeed a classic economic example is that in many economic datasets, price and quantity are positively correlated. Firms set prices higher in high-income cities where consumers buy more; they raise prices in anticipation of times of peak demand. A large body of econometric research seeks to REDUCE the goodness of fit of a model in order to estimate the causal effect of, say, changing prices. If prices and quantities are positively correlated in the data, any model that estimates the true causal effect (quantity goes down if you change price) will not do as good a job fitting the data. The place where the econometric model with a causal estimate would do better is at fitting what happens if the firm actually changes prices at a given point in time—at doing counterfactual predictions when the world changes. Techniques like instrumental variables seek to use only some of the information that is in the data – the “clean” or “exogenous” or “experiment-like” variation in price—sacrificing predictive accuracy in the current environment to learn about a more fundamental relationship that will help make decisions about changing price. This type of model has not received almost any attention in ML.
The answer is interesting, though difficult, throughout. Here are various Susan Athey writings, on machine learning. Here are other Susan Athey answers on Quora, recommended. Here is her answer on whether machine learning is “just prediction.”
Autor, Dorn, and Hanson, on what we know about China
They have a new and excellent summary paper (pdf), and that is Gordon not Robin Hanson:
China’s emergence as a great economic power has induced an epochal shift in patterns of world trade. Simultaneously, it has challenged much of the received empirical wisdom about how labor markets adjust to trade shocks. Alongside the heralded consumer benefits of expanded trade are substantial adjustment costs and distributional consequences. These impacts are most visible in the local labor markets in which the industries exposed to foreign competition are concentrated. Adjustment in local labor markets is remarkably slow, with wages and labor-force participation rates remaining depressed and unemployment rates remaining elevated for at least a full decade after the China trade shock commences. Exposed workers experience greater job churning and reduced lifetime income. At the national level, employment has fallen in U.S. industries more exposed to import competition, as expected, but offsetting employment gains in other industries have yet to materialize. Better understanding when and where trade is costly, and how and why it may be beneficial, are key items on the research agenda for trade and lab or economists.
This is some of the most important work done by economists in the last twenty years.
Facts about business travel
More populous countries have more business travel in both directions, but the volume is less than proportional to their population: a country with 100% more population than another has only about 70% more business travel. This suggests that there are economies of scale in running businesses that favor large countries.
By contrast, a country with a per capita income that is 100% higher than another receives 130% more business travelers and sends 170% more people abroad. This means that business travel tends to grow more than proportionally with the level of development.
While businesspeople travel in order to trade or invest, more than half of international business travel seems to be related to the management of foreign subsidiaries. The global economy is increasingly characterized by global firms, which need to deploy their know-how to their different locations around the world. The data show that there is almost twice the amount of travel from headquarters to subsidiaries as there is in the opposite direction. Exporters also travel twice as much as importers.
That is from Ricardo Hausmann, with further interesting points.
How ideological are economists?
From Ryan Avent:
Anthony Randazzo of the Reason Foundation, a libertarian think-tank, and Jonathan Haidt of New York University recently asked a group of academic economists both moral questions (is it fairer to divide resources equally, or according to effort?) and questions about economics. They found a high correlation between the economists’ views on ethics and on economics. The correlation was not limited to matters of debate—how much governments should intervene to reduce inequality, say—but also encompassed more empirical questions, such as how fiscal austerity affects economies on the ropes. Another study found that, in supposedly empirical research, right-leaning economists discerned more economically damaging effects from increases in taxes than left-leaning ones.
There is considerably more at the link. The Randazzo and Haidt study is from Econ Journal Watch.
China fact of the day
By almost all measures, China’s $3.3 trillion foreign reserves, the world’s largest, look formidable. Except one.
Compared with the amount of yuan sloshing around in the economy, a proxy for potential capital outflows, China’s firepower seems limited. The dollar reserves account for 15.5 percent of M2, a broad measure of money in circulation. That’s the lowest since 2004 and is less than levels in most Asian economies including Thailand, Singapore, Taiwan, Philippines and Malaysia, according to data compiled by Bloomberg.
That is from Ye Xie. Please do note all of the caveats and qualifiers in the longer piece. The claim is not that Chinese reserves are currently in some kind of crisis situation, only that we should not overestimate their import, relative to potential and indeed actual capital outflows.
How to seem telepathic
This is probably one of the most useful things you will learn from MR all year. It is from Maria Konnikova’s new book The Confidence Game:
In 2010, Nicholas Epley and Tal Eyal of Ben-Gurion University published the results of a series of experiments aimed at improving our person and mind perception skills. The title of their paper: “How to Seem Telepathic.” Many of our errors, the researchers found, stem from a basic mismatch between how we analyze ourselves and how we analyze others. When it comes to ourselves, we employ a fine-grained, highly contextualized level of detail. When we think about others, however, we operate at a much higher, more generalized and abstract level. For instance, when answering the same question about ourselves or others — how attractive are you? — we use very different cues. For our own appearance, we think about how our hair is looking that morning, whether we got enough sleep, how well that shirt matches our complexion. For that of others, we form a surface judgment based on overall gist. So, there are two mismatches: we aren’t quite sure how others are seeing us, and we are incorrectly judging how they see themselves.
If, however, we can adjust our level of analysis, we suddenly appear much more intuitive and accurate. In one study, people became more accurate at discerning how others see them when they thought their photograph was going to be evaluated a few months later, as opposed to the same day, while in another, the same accuracy shift happened if they thought a recording they’d made describing themselves would be heard a few months later [TC: recall Robin Hanson’s near vs. far mode]. Suddenly, they were using the same abstract lens that others are likely to use naturally…
Upon reading this passage I realized I have been thinking in these terms for years, without quite realizing it so explicitly.
One implication: if you feel bad one morning, don’t let it get you down and lower your confidence. Other people probably won’t notice your problems.
Another implication: you’ll understand yourself better if, in a given moment, you can pretend to distance yourself from some of your immediate impressions of your day, and treat yourself like a piece of your writing which you set aside for a week so you could look at it fresh.
A third implication is this: you can read other people’s moods better by ignoring some of your overall impressions of them, and by focusing on what they might perceive to be small changes in their situation, appearance, or stress levels.
The original research is here, worth a read (pdf). And here are various reviews of the Konnikova book.
It’s all about women in the South
Wow!! Remember that increasing death rate among middle-aged non-Hispanic whites? It’s all about women in the south (and, to a lesser extent, women in the midwest). Amazing what can be learned just by slicing data.
I don’t have any explanations for this. As I told a reporter the other day, I believe in the division of labor: I try to figure out what’s happening, and I’ll let other people explain why.
That is from Andrew Gelman, there is more at the link.
Anti-GMO Research Under Fire
Nature reports that some of the research most-cited by opponents of genetically modifying crops appears to have been manipulated. In particular, images appear to have been altered and images from one paper appear in another paper describing different experiments with different captions.
Papers that describe harmful effects to animals fed genetically modified (GM) crops are under scrutiny for alleged data manipulation. The leaked findings of an ongoing investigation at the University of Naples in Italy suggest that images in the papers may have been intentionally altered. The leader of the lab that carried out the work there says that there is no substance to this claim.
The papers’ findings run counter to those of numerous safety tests carried out by food and drug agencies around the world, which indicate that there are no dangers associated with eating GM food. But the work has been widely cited on anti-GM websites — and results of the experiments that the papers describe were referenced in an Italian Senate hearing last July on whether the country should allow cultivation of safety-approved GM crops.
Are queens more warlike?
There is a reason chess evolved the way it did:
…we find that queenly reigns participated more in inter-state conflicts, without experiencing more internal conflict. Moreover, the tendency of queens to participate as conflict aggressors varied based on marital status.
Among married monarchs, queens were more likely to participate as attackers than kings. Among unmarried monarchs, queens were more likely to be attacked than kings. These results are consistent with an account in which queens relied on their spouses to manage state affairs, enabling them to pursue more aggressive war policies. Kings, on the other hand, were less inclined to utilize a similar division of labor.
This asymmetry in how queens relied on male spouses and kings relied on female spouses strengthened the relative capacity of queenly reigns, facilitating their greater participation in warfare.
As Chris Blattman tells us, that is from “A new paper, Queens, by Oeindrila Dube and S.P. Harish.”
The truth about Chinese rebalancing toward services
The optimists point to the rise in the share of services in nominal GDP, and the corresponding decline in industrial sectors, as shown in the above left graph. Measured in current prices, the rebalancing appears to be well underway, with the share of industrial sectors falling from 47 per cent in 2011 to 40 per cent now.
However, almost the whole of this rebalancing in nominal terms has occurred because of a large drop in the relative price of industrial products compared to services. In real, inflation adjusted terms (above right graph), there has been no rebalancing whatsoever in the past decade taken as a whole (though there has been a percent or two in 2014-15). The needed shift in real resources – labour and capital – out of the moribund sectors has therefore barely started.
That is from the excellent Gavyn Davies, file under “Ouch.”
Addendum: Scott Sumner comments.
Does it matter that to some extent banks are commonly owned?
There is now a paper on this topic by Azar, Raina, and Schmalz, the main result is this:
We document a secular increase of deposit account maintenance fees and fee thresholds with a new branch-level dataset, as well as substantial cross-sectional variation in these prices and in deposit rate spreads. We then examine whether variation in bank concentration helps explain the variation in prices. The standard measure of concentration, the HHI, is not correlated with any of the outcome variables. A generalized HHI (GHHI) that captures both common ownership (the degree to which banks are commonly owned by the same investors) and cross-ownership (the extent to which banks own shares in each other) is strongly correlated with higher maintenance fees, fee thresholds, and deposit rate spreads. We use the growth of index funds as a source of exogenous variation to establish a causal link from GHHI to higher prices for banking products.
In other words, if companies are owned by the same pension and mutual funds, why should they compete against each other? Imagine managers given financial incentives for greater stability rather than greater risk-taking, so this does not require a publicly traceable conspiracy.
The first paper on this general question was in fact written by me and Ami Glazer about twenty-five years ago, although we never managed to get it published. Our biggest problem was perhaps the lack of clear evidence at the time. This is the best evidence I have seen so far, although I still file this under “speculative”…
For the pointer I thank Uri Bram.
Being drafted during the Vietnam War also hurt your descendents
A decade after their military service, white veterans of the draft were earning about 15 percent less than their peers who didn’t serve, according to studies from MIT economist Josh Angrist.
Now, new research suggests that the draft did more than dim the prospects of that earlier generation: The children of men with unlucky draft numbers are also worse off today. They earn less and are less likely to have jobs, according to a draft of a report from Sarena F. Goodman, an economist with the Federal Reserve Board of Governors, and Adam Isen, an economist at the Treasury Department. (A copy was released by the Fed in December, but research does not reflect the opinions of the government.)
The researchers have not nailed down how, exactly, any of this is happening, nor why the disadvantage appears to be over twice as potent for sons than for daughters. But the work is valuable for showing how the circumstances of one’s parents can have lasting repercussions. This is one way that inequality persists through the generations.
That is from Jeff Guo at Wonkblog.