Sunday, 28 June 2015

More plots on war and culture

Oh this is fun. National identity.

English vs. British:

The two spikes in the 19th century are 1837 and 1859. 1837 is the coronation of Queen Victoria, I suppose. 1859 I'm not sure about.


Here the world war spikes are obvious. Also interesting is the sustained rise from the postwar low. Quite a contrast with both British and English. 

France and America are quite similar. France has huge spikes for the two world wars, and earlier for 1813-15

America has little spikes for the world wars, and possibly a high for Vietnam (1964-1975) though I could not tell you if it would pass statistical significance. (I thought that after Vietnam there would be a postwar low as for Germany. But apparently Vietnam was a tie.)


Friday, 26 June 2015

The impact of war on culture

Here's the occurrence of the noun devoir (meaning duty) in a corpus of French books from 1800-2008:

There are big spikes during major wars. The other place with a noticeable spike is around the 1848 revolution, interestingly.

For comparison I used the verb devoir (to have to). It doesn't have such obvious spikes.

Theory suggests that war could cause to the evolution of human in-group altruism, and evidence shows that conflict indeed makes people more prosocial. This chart shows one channel for that to happen: changes in social norms.

The picture for Germany is not so clear. Germany has two words for duty, Pflicht and Aufgabe. Aufgabe also means "function", so it's more ambiguous.

Here's the graph:

There are spikes in Pflicht around 1871 and 1914-18, but not noticeably for 1939. And for Aufgabe, the big jump comes not in 1939 but in 1933, when the Nazis took over. The early data for Germany also seems more random (though this could relate to the Napoleonic wars, which were a crucible of German nationhood - actually, that's true of France also, but the big early spike in the graph above is around 1820 not 1815).

Lastly the UK - this is the "British English" corpus:

The spikes in the 19th century are around 1830 and 1853. 1830 is a time of civil unrest just before teh passing of the first Reform Bill. 1853 is the Crimean War. 1914-18 shows up as a spike but only at about 1916. And in 1939 there's a wee spike that quickly vanishes... it's easy to fit a post hoc interpretation about the un-idealistic spirit with which Britain approached World War II.

OK, so using the word "duty" in war may not be such a big surprise. What about the most basic group-word, "we"?

Here's France:

Clear spikes during the three major wars.

Here's Germany. (Sorry about the unhelpful x axis labels, blame Google. The lines are 20 year intervals.)

There's a huge spike for WWII and a spike for 1914-18. Actually, the WWII spike is 1945 precisely. A null for 1871. (Earlier on there's something about the 1840s... here my knowledge of European history is lacking!)

Britain is less clear. There are spikes for the two world wars but they stand out less, especially compared to the sustained height of the 19th century:

All of this is very quick and dirty, but I think there is scope for more exploration.

Wednesday, 24 June 2015


Bryan Caplan crystallizes a simple, interesting idea.

Paul Krugman versus the Austerians: the video game.

I don't know whether Paul Krugman is right, but he is a brilliant debater and an example of how one guy can change the discussion.

All we need now is an actual video game.

Monday, 22 June 2015

Linkage on science and freedom

GCHQ uses behavioural science to manipulate the public.

Richard Thaler writes "nudge for good" when he signs his book. Unfortunately, the bad guys have no incentive to follow that advice. It is sad when psychological science is used by unaccountable state securocrats as a source of human weaknesses to exploit.

A book by an academic who works in China saying that meritocracy beats democracy.

As non-Western democracies invest more in education, they are likely to be selective buyers of   research. Social scientists in democracy overwhelmingly favour democracy. It will be interesting to see what happens to social scientists in autocracies.

Saturday, 13 June 2015

Bad arguments about brexit: a compendium

Norway and Switzerland are outside the EU but still have to comply with EU regulations.

Yeah, but we're five times bigger and more powerful than either of these, so we would get a better deal.

Leaving the EU will deduct 2% / add 1% to our GDP.

Anyone estimating this is making an educated guess. If the confidence intervals around the estimate aren't big enough to make it useless as a guide to policy, then the author is kidding us and/or herself.

We have to contribute billions to the EU budget.

We don't know how much the economic benefits/costs will be (see above), but we do know that compared to them, this is peanuts.

The EU has kept the peace in Europe.

By being a Franco-German pact, maybe. Our exit is unlikely to change the nature of that pact, and I doubt that we will be planning war on Berlin.


I'll try to keep this updated. My view is that "expert" arguments have relatively little to contribute to this question: the choice is an existential one, not a technical one. It's not that facts are irrelevant, but the important effects of brexit will be political, not economic, and in predicting politics, good judgment is more important than technical expertise. Full disclosure: I intend to vote to stay in the EU, though I can imagine changing my mind.

The next half-century is likely to be very bloody

Most researchers – if not quite a consensus – think the world is getting more peaceful. Steven Pinker's book puts this view in its broadest form. I am not an expert on violent conflict, but I disagree. (I don't feel bad about expressing my amateur's opinion. Experts on violent conflict, unlike, say, experts on quantum physics, are not necessarily much better able to make predictions than the rest of us, and the wisest of them would admit that.)

I expect the next half-century to have more, and perhaps bigger, wars, massacres and genocides than the first half of the twentieth century.

My prediction is based on a simple two step theory of modern history:
  1. Economic growth makes people rich;
  2. Then they fight over the spoils.
In the 19th century, Europe got rich. European nation-states then fell to fighting over the spoils of empire. This led to two world wars, which some historians redescribe as a single "thirty years' war". There were also many episodes of violence within states, including genocide.

In the late 20th century, the whole world started to get rich. This means there will be bigger fights.

Specifically, as the US gets less powerful, there will be a period without a "hegemon", a world policeman. When the teacher leaves class, the children start to misbehave.

Technological progress also enables people to kill each other faster, when they want to, and this is likely to make some conflicts very bloody.

There are three main risks of large-scale violence.
  • Big interstate wars between growing powers, or between rising and status quo powers. So, US-China or China-India.
The risk of these is low because these countries have nuclear weapons, which makes war very costly, and they are also large and complex enough to have reasonably sane leadership and bureaucracies, which lowers the risk of them doing something stupid. On the other hand, if it happened, it would be catastrophic.
  • Middle-sized interstate wars, or large-scale civil wars in large countries. 
We have this now in the Middle East: part of the reason is surely the US snoozing on the job after its unfortunate adventure in Iraq. When the countries involved have large populations, and war is not limited or localized, the number of casualties can get very large. There is not much reason to think that nuclear weapons will make conflict less likely, if they are in the hands of incompetents or madmen. (Remember they have been used twice already.)
  • Massacres and genocide.
Economic growth benefits some groups more than others, and as Amy Chua and before that Donald Horowitz pointed out, this can cause violence. There are many vulnerable groups in the developing world. Probably the single biggest one is India's Muslims. These are a minority; there is a well-established political party which uses anti-Muslim hate to get ahead, and which is now running the country; and there is a history of violence between Muslims and Hindus, including the million killed during Partition. Also, the vulnerable population is huge – 180 million.
There is no deep social science insight or data analysis behind my belief. (The 2 step theory is influenced by the book Violence and Social Orders, from the economic historian Douglas North and his coauthors. Their theory is much subtler than my dumbed-down version, though.) It is just a personal view. I am not very sure of it, because I am aware of my profound ignorance, and of my many previous mistaken predictions.

Still, I would like other people to worry about it more: I hope the experts will adduce better reasons to show I am right or wrong. This is not because we can do much to stop mass violence, but because it would be wise to prepare for it.

There is a third step of history: after people have fought to a standstill, they settle down and continue to get rich. This seems to have happened in the developed West. I hope it will happen in the whole world. Until then the ride will probably be bumpy.

Thursday, 11 June 2015


How community groups helped implement genocide in Rwanda.
Every Saturday before 1994, Rwandan villagers had to meet to work on community infrastructure, a practice called Umuganda. This practice was highly politicized and, in the years before the genocide, regularly used for spreading political propaganda. To establish causality, we exploit cross-sectional variation in meeting intensity induced by exogenous weather fluctuations....

Sunday, 7 June 2015

R tip: bubble plot

To draw a bubble plot for data with many identical values:

sizeplot(x, y, pch = 19, col = "green")

The area of the bubble is proportional to the number of values at that point.

Wednesday, 3 June 2015

Tuesday, 2 June 2015

Animated map of police shootings

I wrote a simple R script to create an animated map of police shootings this year.

R packages used: RgoogleMaps to get latitude and longitude, lubridate for handling dates, anim.plots to create the animation, and maps for the US map.

Data from the Guardian story.

dfr <- read.csv("~/Downloads/thecounted-data/the-counted.csv")
latlon <- sapply(paste(dfr$address, dfr$city, dfr$state,  sep = ","), getGeoCode)
dfr$lat <- latlon[1,]
dfr$lon <- latlon[2,]

dfr$race2 <- as.character(dfr$raceethnicity)
dfr$race2[! dfr$race2 %in% c("Black", "White", "Hispanic/Latino")] <- "Other"
sym.cols <- AddAlpha(c("brown4", "orange", "grey", "pink3"), .8)
dfr$racecol <- sym.cols[factor(dfr$race2)]
dfr$time <- dmy(paste(dfr$day, dfr$month, dfr$year))

map("state", ylim = c(20, 49))
legend(-125, 25, col=sym.cols, legend = c("Black", "Hispanic", "Other", "White"),
      pch = 19, horiz = TRUE, bty="n")
legend(-125, 23, pch = c(2, 1), legend = c("Armed", "Unarmed"), horiz = TRUE, bty = "n")
anim.points(dfr$lon, dfr$lat, times = dfr$time, col = dfr$racecol, pch = ifelse(dfr$armed=="Yes", 17, 19))