2015-10-22 War and Statistics

I recently saw a Tweet by William Easterly, citing work by Max Roser, claiming that global deaths in conflicts are trending downwards (see pic below). Roser had tweeted it before but I could not find it on his website, *Our World in Data*. On Twitter, Sébastien Broos posted a link to a paper by Pasquale Cirillo and Nassim Nicholas Taleb (”Black Swan”): On the tail risk of violent conflict and its underestimation.

by William Easterly

tweeted it before

On the tail risk of violent conflict and its underestimation

The abstract: *We examine all possible statistical pictures of violent conflicts over common era history with a focus on dealing with incompleteness and unreliability of data. We apply methods from extreme value theory on log-transformed data to remove compact support, then, owing to the boundedness of maximum casualties, retransform the data and derive expected means. We find the estimated mean likely to be at least three times larger than the sample mean, meaning severe underestimation of the severity of conflicts from naive observation. We check for robustness by sampling between high and low estimates and jackknifing the data. We study inter-arrival times between tail events and find (first-order) memorylessless of events. The statistical pictures obtained are at variance with the claims about “long peace”.*

They repeat this in the summary at the end: *In other words, a large event and even a rise in observed mean violence would not be inconsistent with statistical properties, meaning it would justify a “nothing has changed” reaction.*

That’s harsh. It means that statistics do not support our subjective point of view that the world has been more peaceful in recent decades. Statistics tell us that both the current lull and a threefold increase are both plausible. There is no rest for the wicked.

The first two and the last two pages are the most readable pages. I’m not big into statistics so I can’t comment on the actual math. As a listener to the Chinese History Podcast and having just heard about *The Rise and Fall of the Qin* (Part 1, Part 2), and all those battles with 400,000 enemy soldiers buried alive after the fighting was over... Ugh! I looked at that big red bubble for *An Lushan* on page 2 for quite a while. There is a short discussion about the numbers on Wikipedia. Some people say that the decrease is due to the census system breaking down and due to the shrinking territory. The raw numbers, however, show the last census before the rebellion recording a population of 53M in 755 and of 17M in 764, the year following the end of the rebellion. That would mean 36M dead, or about *one sixth of the world population at the time*.

Chinese History Podcast

Part 1

Part 2

400,000 enemy soldiers buried alive

a short discussion about the numbers on Wikipedia

Now consider one of the methods Cirillo and Taleb used to make their risk assessment: *a log transformation to account for the fact that the number of casualties in a conflict cannot be larger than the world population.*

Also consider that their examination showed *no visible autocorrelation*. The large wars are independent from each other. There is no “slowing down” or “speeding up” of occurrences to observe.

Anyway, what a chilling thought to start the day with. 🙁

​#War