I just finished the second of two posts for GiveWell on the heated academic controversy over whether it is a good idea to mass-deworm children in regions where the parasite infections are common. The first post focusses on the “internal validity” of a particularly influential study that took place along Lake Victoria, in Kenya, in the late 1990s. The second thinks through how safely we can generalize from that study to other times and places. It has a lot more graphs, including some that look pretty wormy…
Most econometric methods are buttressed by mathematical proofs buried somewhere in academic journals that the methods converge to perfect reliability as sample size goes to infinity. Most arguments in econometrics are over how best to proceed when your data put you very far from the theoretical ideal. Prime examples are when your data are clustered (some villages get bednets and some don’t) and there are few clusters; and when instruments are weak (people offered microcredit were only slightly more likely to take it).
Mucking about in such debates recently, as they pertain to criminal justice studies I’m reviewing, I felt an urge to get back to basics, by which I mean to better understand the mathematics of methods such as LIML. That led me back to linear algebra. So I’ve been trying to develop stronger intuitions about such things as: how a square matrices can have two meanings (a set of basis vectors for a linear space, and the variances and covariances of a set of vectors); and what the determinant really is.
I have a new post on openphilanthropy.org suggesting that there was indeed an urban crime wave in the US in the last couple of years, but that it was mainly restricted to homicide and assault with a firearm, and may well have peaked last year.
I started studying the causes and consequences of incarceration for the Open Philanthropy Project. The subject is full of mysteries. Here’s one.
As best we can measure, the US crime rate rose from the mid-1960s to the early 1990s and then reversed:
(Following FBI definitions, this graph is of “Part I” crimes and excludes excludes drug crime, white collar crime, drunk driving offenses, traffic violations, and other minor crimes. The property crime rate is graphed against the left axis, the violent crime rate against the right.)
The strange thing is, the experts aren’t completely sure why the rise and fall. More…
In 1980, Fidel Castro suddenly allowed thousands of Cubans to leave the country—if they could find a way out. Americans, many of Cuban extraction, swooped to the rescue by bringing lots of boats to the Cuban Port of Mariel. It was called the Mariel boatlift. Some 125,000 Cubans moved to America in a matter of months and perhaps half settled in Miami.
Some 10 years later, economist David Card viewed the Mariel boatlift as a natural experiment and used it to study how immigration affects wages and employment in the receiving country. He concluded there was not much discernible impact in Miami. His paper is seminal, both for its counterintuitive finding and for its introduction of the natural-experiment approach to the study of immigration’s impacts.
Last month, George Borjas, an economist and Cuban emigré himself, revisited the data and came to opposite to conclusion from Card’s. The boatlift hurt the wages of low-education Miamians.
So I dug into the data. Borjas’s work ended up not convincing me. More on the GiveWell blog.
The fourth installment in my series on geomagnetic storms is now up on the GiveWell blog.
The first three posts were about the odds of a big storming hitting earth. This one shifts to the question of likely impacts. Mostly I think the evidence is reassuring. But given the stakes, I think we should not relax, and instead support more thorough research.
The Open Philanthropy Project just released another big literature review of mine, this one on whether taxing alcohol save lives. I conclude that it probably does.
That’s hardly shocking. Making stuff more expensive generally leads to people to buy less. And alcohol in excess is bad for you. Perhaps the more significant finding is about the number of lives that could be saved, which is not so great next to other things that “Open Phil” might fund. E.g., we dream of financing the invention of a new research technique that leads to a cure for Alzheimer’s. The grant chasing that dream would be a longshot…but then so might be funding advocacy for raising taxes.
Coroners in the US attribute 23,000 deaths/year to alcohol-caused diseases, according to my calculation (see the report for more). The most rigorous studies I found produced a rather wide range of elasticities of death rates with respect to alcohol prices: 1–3. That means each 1% price rise reduces deaths 1–3%. And, if you do the math carefully, tax hikes sufficient to raise alcohol prices 10% would cut the alcohol death rate 9–25%, or 2,000–6,000 lives/year. This math leaves out any reduction in deaths from drunk driving, which currently amount to 10,000/year. The benefit there would presumably be of the same order of magnitude.
A few interesting things I learned and did along the way:
My third GiveWell post on geomagnetic storms is up. This one marshalls a branch of statistics called Extreme Value Theory, which I learned about through this work. Last year I wrote a program to bring EVT techniques to Stata.
The second post in my series on geomagnetic storms is up on GiveWell.org. It is arguably the most important and interesting in the series. It explains why I think past storms, reaching back to 1859, were probably at most twice as strong as anything our electricity-dependent societies have experienced in recent decades—and shrugged off.
Do you remember the great storms of 1982 and 2003? I didn’t notice them either. And probably you survived the Québec blackout of 1989, which was mostly over within 11 hours. Yet maybe that last doubling in storm intensity would inflict far, far more than twice as much destruction on the grid. Or maybe the grid has become much more vulnerable since 1989, even though grid operators have learned from that experience. It’s also possible I’m wrong that doubling is the worst we should fear. For all these reasons, I still think the threat deserves more attention from researchers, industry, and governments.
As I mentioned in my previous post, the strongest proponent for the view that the worst case is much worse, is John Kappenman, who has argued for a multiplier of 10 rather than 2. In the new post and the report, I trace this number in part to an obscure book of scientific scholarship written in 1925 by a Swedish telegraph engineer in French. The search involved talking to an electrical engineer in Finland, people at the Encyclopedia Britannica in Chicago (who were very helpful), and ordering said obscure book from a German book shop. Author David Stenquist describes how the storm of 1921 caused copper wires running into a telegraph office to melt—but not iron ones. He deduces that the storm-induced voltage on the line could not have been as high as 20 volts/kilometer. Yet through a scholarly game of telephone over the decades, this observation got turned on its head.
My long-promised report for the Open Philanthropy Project on geomagnetic storms is posted. (Data, code, and spreadsheets are here.) The first of a series of posts based on the report just appeared on the GiveWell blog.
This has been one of the most fun projects I’ve worked on because it slices across so many disciplines, from statistics to power engineering to astrophysics. My grasp of those subjects declines in the order listed…but I think I learned enough to reach a preliminary assessment.
The risk that a major solar cataclysm could so disrupt the earth’s magnetic field as to deprive continent-scale regions of power for years looks low to me—lower than the most attention-getting voices, almost by definition, have suggested (Pete Riley, John Kappenman). Nevertheless, a long-term, large-area blackout would do so much harm, and the risk is so poorly studied, that it absolutely deserves more attention from researchers, industry, government, and philanthropies. My preliminary risk assessment could be wrong.
I just discovered that an elite, independent scientific advisory group for the US government arrived at a similar conclusion in 2011.
It follows that the most emphatic analysts, even if they have overshot, have done a service by drawing attention to the issue. This is for me a familiar paradox.
After I blogged Cirillo and Taleb’s new paper on the long-term trend in war deaths, I read other commentaries on the debate (William Briggs, Dart-Throwing Chimp, STATS.org) and interacted with the authors. All that sharpened my thinking. Refinements:
- The paper is postured as a rejoinder to Steven Pinker. But I think if you are going use statistics to show that someone else is wrong, you should 1) state precisely what view you question, 2) provide examples of your opponent espousing this view, and 3) run statistical tests specified to test this view. Cirillo and Taleb skip the first two and hardly do the third. The “long peace” hypothesis is never precisely defined; Pinker’s work appears only in some orphan footnotes; the clear meaning of the “long peace”—a break with the past in 1945—is never directly tested for.
As promised, I have posted a database of variables that I developed for my analyses of the contributions of loans to aid. Also now available is an SQL Server database with the underlying data and logic (You can download the needed software, SQL Server Express, for free but you need to know what you’re doing.)
This database includes a construction of the new DAC ODA variable—which I think you can’t get anywhere else—and my preferred alternatives based on DDRs. All are back-calculated to 1980.
I have also posted a new version of my ODA redefinition paper, “Straightening the Measuring Stick,” which updates last year’s CGD working paper. I presented the newer one last month in Geneva at a conference jointly organized by the IMF and the Graduate Institute Geneva. The proceedings of the conference, including my paper, are now in process with a journal. The largest changes were necessitated by the Development Assistance Committee rule changes agreed in December: rather awkwardly, I had to switch my theme from what the DAC should do to what it had and ought to have done. I blogged the most interesting updates to my thinking in March.
Two intellectual titans are arguing over whether humanity has become less violent. In his 2011 book, Steven Pinker contends that violence is way down since the stone age, or even since the Middle Ages. He looks at murder, war, capital punishment, even violence against animals.
But in a working paper released yesterday, Pasquale Cirillo and Nassim Taleb, the latter the author of The Black Swan: The Impact of the Highly Improbable, contend that Pinker has it wrong. Well, more precisely (lest I incite a riot with demagoguery) they challenge the notion that the great powers have enjoyed a distinctly long peace since World War II.
This paper launches a second round in a war of words (and equations) over the long peace. Previously, Taleb said the The “Long Peace” is a Statistical Illusion. Pinker suggested Taleb was Fooled by Belligerence.
Everything I know about the history of violence, I learned from Pinker’s TED talk. And I have not mastered the mathematical methods marshalled by Cirillo and Taleb. But I do know something about them, having written a program to do them. I think Taleb’s latest salvo mostly misses its mark. Here’s why:
I’ve been reviewing two questions in depth for the Open Philanthropy Project, my 75% employer: How much should we worry about geomagnetic storms? And do alcohol taxes save lives? I hope to share drafts on both soon.
In the meantime, I want to share what is for me a surprising discovery, assuming it is true. The idea that moderate drinking is better for your heart than abstention looks headed for the ash heap of history like so many upended lessons from observational epidemiology. (Years ago, I blogged a certain example of this trend, relating to hormone replacement therapy for post-menopausal women: roughly speaking, observational studies said it was good; randomized trials showed it was not just not good, but bad.)
Here’s a chunk from the draft text that explains how this issue relates to whether alcohol taxes save lives (net), and how I reached my current understanding. I emphasize that I based this write-up on a day or so of reading. That said, my priors about the reliability of studies of various types come from longer experience. I’d welcome critical reactions, and sharing of this post in order to provoke them.
Last year I blogged much about how the Development Assistance Committee (DAC) should revise the definition of Official Development Assistance, particularly in its treatment of loans. Then in December, the DAC members made their decision. Some changes I advocated (though hardly with originality), entered the revised definition of ODA. Others did not. On the day of the announcement I blogged my reactions and some preliminary analysis.
Now I am revisiting this matter as I prepare for an IMF-organized conference next month in Geneva.
Using historical data, I have done my best to apply the new ODA loan rules to historical data. So I can compare ODA computed the new way to ODA computed according to the old way, and to ODA computed in a couple of ways I prefer.
To recap, the DAC decided to:
- The domestic economic impacts of immigration
- The impact of life-saving interventions on fertility
- Are the benefits of moderate drinking a myth?
- Malaria maps, then and now
- Stand up for your health
- Of Technocrats and Autocrats: Review of Bill Easterly’s Tyranny of Experts
- What’s the best way to count loans as aid?
- Undue credit: Are France, Germany, and Japan subverting the definition of aid?