A couple of weeks ago (here), I used the blog to launch an experiment in crowdsourcing assessments of coup risk for 2015 by way of a pairwise wiki survey. The survey is still open and will stay that way until the end of the year, but with nearly 2,700 pairwise votes already cast, I thought it was good time to take stock of the results so far.
Before discussing those results, though, let me say thank you to all the people who voted in the survey or shared the link. These data don’t materialize from thin air. They only exist because busy people contributed their knowledge and time, and I really appreciate all of those contributions.
Okay, so, what does that self-assembled crowd think about relative risks of coup attempts in 2015? The figure below maps the country scores produced from the votes cast so far. Darker grey indicates higher risk. PLEASE NOTE: Those scores fall on a 0–100 scale, but they are not estimated probabilities of a coup attempt. Instead, they are only measures of relative risk, because that’s all we can get from a pairwise wiki survey. Coup attempts are rare events—in most recent years, we’ve seen fewer than a handful of them worldwide—so the safe bet for nearly every country every year is that there won’t be any coup attempts this year.
Smaller countries can be hard to find on that map, and small differences in scores can be hard to discern, so I also like to have a list of the results to peruse. Here’s a dot plot with countries in descending order by model score. (It’d be nice to make this table sortable so you could also look for countries alphabetically, but my Internet fu is not up to that task.)
This survey is open to the public, and participants may cast as many votes as they like in as many sessions as they like. The scores summarized above come from nearly 2,700 votes cast between the morning of January 3, when I published the blog post about the survey, and the morning of January 14, when I downloaded a report on the current results. At present, this blog has a few thousand followers on Wordpress and a few hundred email subscribers. I also publicized the survey twice on Twitter, where I have approximately 6,000 followers: once when I published the initial blog post, and again on January 13. As the plot below shows, participation spiked around both of those pushes and was low otherwise.
The survey instrument does not collect identifying information about participants, so it is impossible to describe the make-up of the crowd. What we do know is that those votes came from about 100 unique user sessions. Some people probably participated more than once—I know that I cast a dozen or so votes on a few occasions—so 100 unique sessions probably works out to something like 80 or 90 individuals. But that’s a guess.
We also know that those votes came from lots of different parts of the world. As the map below shows, most of the votes came from the U.S., Europe, and Australia, but there were also pockets of activity in the Middle East (especially Israel), Latin America (Brazil and Argentina), Africa (Cote d’Ivoire and Rwanda), and Asia (Thailand and Bangladesh).
I’ll talk a little more about the substance of these results when I publish my statistical assessments of coup risk for 2015, hopefully in the next week or so. Meanwhile, number-crunchers can get a .csv with the data used to generate the map and table in this post from my Google Drive (here) and the R script from GitHub (here). If you’re interested in seeing the raw vote-level data from which those scores were generated, drop me a line.