Political scientist Jay Ulfelder, who’s quickly becoming the Nate Silver of coups, has accurately predicted another one with the ouster of President Francois Bozizé of the Central African Republic. On Monday, rebel fighters cemented control over the country’s capital, Bangui, as Bozizé fled to Cameroon for temporary refuge.
A coup isn’t something anyone really wants to rejoice in predicting, but in Ulfelder’s 2013 forecast of coups, the Central African Republic was ranked in the top 20 most at-risk countries. (Not bad for a phenomenon that’s famously difficult to predict.) This comes just one year after Ulfelder successfully predicted coups in Mali and Guinea-Bissau, which he had also categorized as "high-risk" countries.
Now Ulfelder, ever the humble academic, noted that his forecast "presaged" the events in Bangui, but he’s not claiming an outright victory because his working definition of coup doesn’t include rebel victories over a government (military coups are more of what he’s thinking of, though other definitions certainly would identify this weekend’s events as a coup.)
In any event, given his strong track record, it’s worth re-examining what he wrote last summer in Foreign Policy about what makes a country "at-risk":
Most countries in the top 20 land there because they are poor and have competitive authoritarian or partially democratic political regimes. Unsurprisingly, coups also turn out to be a recurrent problem; the risk is higher in countries that have experienced other coup attempts in the past several years, a factor common to the top eight countries on this list. Active insurgencies also increase the risk of a coup, and this factor affected the 2012 forecast for countries like Ethiopia, Mali, and Sudan. Ditto for civil wars and popular uprisings in regional neighbors and slow economic growth, common themes in several regions, including West and Central Africa.
Of course, the top of the list isn’t the only part of it that deserves our attention. Statistical forecasts can also be useful when they surprise us in the other direction, suggesting that some risks aren’t as serious as we thought.
Using statistics to forecast political events isn’t as complicated as it sounds. At its core, it’s all about recognizing the right patterns. Generally speaking, you start by building a list of similar events in relevant cases in the past; then you assemble data on likely risk factors; next, you use statistical techniques to generate an algorithm that captures useful patterns in those data; and, finally, you apply that algorithm to current data to get a forecast.
Well, it’s worked all right thus far.